PATH News & Signals

meme11h agomeme

@RWA_Inc_ @centrifuge Agreed. Repo works because everyone knows who can seize what and how fast. Writing that clarity into the token is what lets RWAs work as real collateral. Reply to @RWA_Inc_: @centrifuge the collateral isn't the hard part, the liquidation path is. repo works because everyone knows who can seize what and how fast. onchain lending needs that same clarity written into the token

35/100·CLong

meme12h agomeme

Ziren is coming to Metis.👏👏👏 https://x.com/kevinliu/status/2117789246006133026 Quoted @kevinliu: Our team has taken the hard path on every decision for many years. Build a zkVM from scratch on MIPS. A working decentralized sequencer set. A truly trust-minimized Bitcoin bridge. Post-quantum security in progress across the stack. And now real-time proving on Ethereum. Ziren is already the heart of GOAT Network, and work is now underway to bring it to Metis.

55/100·BLong+4

meme13h agomeme

The Base ecosystem ships hard Here are 25 things that launched and got announced in October 1) @mastercard launched Agent Pay for Machines with Cloudflare. Enabling AI agents to pay for digital services using x402 on Base 2) @cloudflare opened its Monetization Gateway on Base. Letting APIs charge agents per request with settlement in USDC 3) @bitso launched Bóvedas on Base. Enabling users access to DeFi tools on MXNB and USDC, powered by @morpho and curated by @gauntlet_xyz 4) @pendle_fi launched PendleLoop on Base. Users can loop long Coinbase Tokenized Stocks on @aave markets and claim incentives on @merkl_xyz 5) @onepay now lets users send USDC straight into their Checking balance on Base. People who get paid in stablecoins can spend or send that money without first routing it through an exchange 6) @ripple launched RLUSD on Base. Ripple is providing liquidity so users can hold and move the stablecoin on Base 7) @openstandard launched OUSD on Base. Users get 1:1 USD conversion paths through Coinbase, Mastercard, Stripe, and Visa integrations 8) @itstuyo rebuilt its entire app. The new Tuyo matches bigger ambitions for banking, with a full redesign now live on Base 9) @dyli_io launched on Base. Collectors can rip packs, trade collectibles, vault favorites or have them shipped, with a refreshed pack experience now live 10) @thunespayments is integrating stablecoins across its global payout network on Base. Letting businesses send and receive cross-border stablecoin payouts 11) @marqeta is bringing stablecoin-backed cards to enterprise customers, with settlement on Base. Giving companies a way to issue cards funded by stablecoins for spending 12) @atomhq now supports stablecoin checkout and seller payouts on Base. Users can now access over 2 million domains with stablecoins 13) @smbcnikko_jp is building an institutional DeFi Gateway for Japan on Base. Opening a regulated path for Japanese institutions into DeFi 14) Yuno added stablecoin payments acceptance on Base. Merchants can enable one-click stablecoin checkout without new rails or extra integration work 15) @superformxyz brought SuperNVDA and SuperTSLA to Pendle on Base. Holders can lock in fixed yield, earn rewards, or position for yield upside on tokenized stocks 16) @birchhill_io launched the Groma Ecosystem Vault on Base. Users can lend USDC backed by Boston and Providence apartment buildings 17) @falconxglobal's USDC credit facility is live on Base. Lenders deposit native Base USDC to generate returns from institutional credit 18) Gennius is bringing digital-asset services to LatAm banks with Coinbase, starting in Argentina. Banks can convert eligible deposits into ONED USD, transfer 24/7, and enable Visa spending 19) @torosfinance launched Toros Options on Base. Users can trade over BTC, ETH and more asset options with cashback, and positions are tokenized on Base 20) @exa_app launched the first credit card backed by Coinbase Tokenized Stocks. Enabling users link a Coinbase account and spend against tokenized stock collateral 21) @lucamoneyapp extended its app to tokenized stocks on Base. Letting people in LATAM access them through the same simple digital-dollar experience 22) @zothdotio launched zMAG7 on Base. One stablecoin deposit holds equal-weight AMZNc, AAPLc, NVDAc, GOOGLc, METAc, MSFTc and TSLAc, backed by real shares in regulated custody 23) @mochatradeapp received its full $500K YC investment in USDC on Base 24) @credifi launched Credifi Zero. Anons and agents can qualify for credit with zero personal info 25) @opensea opened OpenSea Mobile to everyone. With new trade options, including memes, art, trading cards and perps on Base from one app Retweeted by @tn_pendle from @base

35/100·CLong+19

news11/3news

29 October Binance Security Incident: Initial Findings and Response Update

On 29 October 2026, Binance experienced a serious and highly sophisticated security incident. This should not have happened, and we apologise for the concern, disruption and uncertainty it caused our users, partners and the wider industry.
Our first responsibility is to protect users. Binance will absorb the financial impact of the stolen funds, not our users. We have already set aside more than 100% of the funds required to cover the loss. As of 31 October 2026, 98% of the impacted token assets have been replenished and are verifiable on our Proof of Reserves page. We are in the process of purchasing the outstanding, more illiquid assets.
Approximately USD 6.34 billion in digital assets were impacted by the incident. As of this update, approximately USD 2.85 billion in digital assets have been frozen or intercepted, subject to reconciliation and market price volatility.
We have taken immediate action to contain the incident and strengthen the systems involved. The entry points identified as having been used in the attack have been closed. Systems touched in the attack path have been rebuilt or replaced. We have also implemented tighter device controls, enhanced wallet-service risk controls and expanded AI-based behavioural anomaly detection.
This is our initial report. It covers what we know today and what we have already improved. The investigation remains ongoing and our findings will be subject to further independent review. Some technical details will remain confidential to prevent malicious actors from using them to plan attacks. We are engaging independent security firms for further review.
The latest information released by the Federal Bureau of Investigation indicates that the Democratic Peopleâ€™s Republic of Korea is responsible for this incident.
We are sharing this initial report because security improves when the industry learns together.
What happened on 29 October
The attack unfolded quickly across several networks.
The attacker exploited unauthorized access to an internal wallet signing service to initiate transactions outside the normal withdrawal process.
Between 20:05 UTC and 20:09 UTC, five Binance hot wallets sent 233.46 million XRP to an attacker-controlled address.
Binance hot wallets on Solana and Optimism sent approximately 970,500 SOL and 44.2 million OP to attacker-controlled addresses.
Binance then received an internal alert regarding large-volume transfers and immediately began investigating.
Binance hot wallet sent 18,270 BTC to an attacker-controlled address.
The attacker then used the signing service to exploit Ethereumâ€™s EIP-7702 authorization mechanism. EIP-7702 allows an account to delegate certain execution rights to a smart contract without exposing the private key. In this case, the attacker used that mechanism to give an attacker-controlled smart contract partial execution rights over a Binance hot wallet.
Binance immediately initiated multiple containment measures, including suspending withdrawals to restrict further outflows, coordinating with partners to freeze affected tokens, and taking steps to revoke unauthorized permissions. These efforts froze a total of $2.85 billion in assets.
However, suspending withdrawals alone could not stop all unauthorized asset movements. The malicious EIP-7702 authorizations existed on-chain and operated independently of Binanceâ€™s normal withdrawal infrastructure.The attackers submitted 50 signing transactions to manipulate the account nonce and establish malicious delegation authorizations. Once established, these authorizations enabled the attackers to continue moving assets independently of Binance's normal withdrawal infrastructure. Binance needed additional transactions to revoke these on-chain authorizations.
The containment process involved additional technical complexities. Because the hot wallet used Multi-Party Computation (MPC), there was no single private key that could be used to immediately revoke the malicious EIP-7702 authorizations. The team therefore developed and executed a specialized procedure to revoke these authorizations, which required additional time. Furthermore, suspending withdrawals and shutting down hot wallet services did not automatically invalidate existing on-chain authorizations.
These challenges extended the time required to fully contain the unauthorized asset movements. We are reviewing the incident response process, signing policy coverage, hot wallet controls and identifying opportunities to further strengthen our ability to respond rapidly to threats in the future.
The final unauthorized transfer of 2,961.9 ETH finalised on-chain at 23:56 UTC. The transaction had previously been broadcast and remained pending before subsequently finalising. No further unauthorised transfers were observed after.
At 00:10 UTC on 30 October 2026, Binance executed a new authorization from the Binance14 hot wallet to the zero address, clearing the EIP-7702 delegation code.
Attack Preparation
Our current investigation indicates that this was a sophisticated, multi-stage campaign. It was not a single isolated breach or a simple phishing incident. The attackers combined a range of techniques, including social engineering by fake recruitment interviews, malicious software downloads, compromised internal AI tooling, endpoint compromise, and an advanced transaction-authorisation method.
June 2026 â€” Supply chain attack. Engineer A from Wallet DevOps downloaded a tampered Homebrew installation in order to install a popular AI coding tool after searching for it on a search engine. Our security team detected the anomaly, replaced the device, and rotated all related credentials as a proactive security action. However, Engineer Aâ€™s role and his related colleagues on our wallet infrastructure team was exposed to the hacker.
Throughout 2026, Binance's security team detected and prevented 57 suspected North Korea-linked cyber threat cases. These attempts reflect the persistent threat posed by state-sponsored actors targeting the cryptocurrency industry. The techniques include fraudulent job interviews, malicious software disguised as popular developer tools, compromised software packages and browser extensions, fake video conferencing applications, and social engineering campaigns designed to steal credentials or deploy remote access malware.
September â€” social engineering. Software Engineer B was contacted through a legitimate job-hunting platform by an individual posing as a Project Director. Software Engineer B had a profile on the platform that referenced his experience at Binance. The purported Project Director arranged a first-round interview with a HR representative, followed by a second-round interview with an individual claiming to be the CEO of a fictitious company.
In the first round, the purported HR representative conducted a live video interview with Software Engineer B. The interviewer shared a social media account with him, including personal photos and videos posted on the platform, in order to build trust. This was more advanced and deceptive than previously seen social engineering approaches.
The second interview was conducted in a popular social chat app, near the end of the interview, Engineer B was instructed to download an â€œassignmentâ€� containing a malicious code repository onto his work computer. The repository contained malicious code that executed during the setup process, initiating a series of unauthorized activities.
The malware employed sophisticated obfuscation and evasion techniques, including replacing a legitimate development tool component with a malicious version to execute commands and establish communication with an attacker-controlled external server. The 0-day malware was not detected by the endpoint security protection tools at the time. The attacker also took steps to conceal the malicious activity and remove traces of the compromise, making detection and subsequent investigation more challenging.
The script executed the below steps by applying obfuscation techniques to evade security detection.
CREATE /opt/homebrew/Cellar/pyenv/2.8.0/libexec/pyenv_
MODIFY pyenv_ â†’ 3,414 B
DELETE .../libexec/pyenv (3,344 B) â†� remove the valid pyenv
RENAME pyenv_ â†’ pyenv (3,414 B) â†� replace it with a malicious version
DELETE /opt/homebrew/bin/pyenv (3,414 B)
ln (create a new symlink) + spawn zsh
20 October â€” use of an internal AI tool as a springboard. The attacker leveraged the compromised computer of engineer B and used an internal AI tool to gain control over the Wallet DevOps Engineer A and another Wallet Software Engineer C. The attacker soon discovered Wallet DevOps Engineer A had a container deployment permission token on his computer.
29 October â€” reach into wallet signing operations. On 29 October between 17:15 and 19:56, the attacker logged onto Software Engineer Câ€™s computer, used the credentials stolen from Wallet DevOps Engineer A, deployed a container, and triggered a malicious script to launch the attack.
What we have done
As of Nov 3, 2026, we have taken the following actions:
Rebuilt the affected wallet system in a new environment.
Rebuilt the affected internal AI tool and access controls.
Replaced affected devices.
Decommissioned credentials and tokens related to affected engineers.
Completed enhanced security scans across all employee devices.
Enhanced AI-based behavioural anomaly detection.
Strengthened wallet-service risk controls.
Worked with law enforcement, blockchain intelligence firms and industry partners to trace and freeze affected assets.
We will continue to strengthen our measures as the investigation progresses. There are a number of security measures we have implemented and are implementing that will remain private to prevent malicious actors from compromising our security.
Further actions
We are taking the following steps immediately.
Independent cybersecurity review
We are engaging independent third-party cybersecurity experts to conduct a comprehensive review of the incident, validate our findings and recommend additional measures. The scope and findings of that review will inform our continuing remediation work.
Expanded investment in AI security and prevention
We are increasing investment in AI-enabled detection, endpoint protection, developer-environment security and behavioural anomaly monitoring.
Security research fund and industry initiatives
We are working on establishing a security research fund and targeted initiatives with leading AI companies, industry participants and security firms focused on AI security in blockchain environments.
We will continue to proactively share information as and when we have it.

78/100·AShort+10

news11/3news

Why Quantum Security Is Not Enough to Secure Your Blockchain For decades, the quantum question was when a quantum computer may arrive. In 2026 a second question took over: how fast is the math moving? AI is now helping find cheaper attacks and new proofs at a pace human research alone has not matched. That shortens the road to Q-Day, the day a quantum computer can break today’s cryptography. It also raises a new risk: some cryptography may fall to better math on ordinary computers, with no quantum computer involved. How is AI speeding up attacks on cryptography? AI is making attacks on cryptography faster and cheaper to find, every few months. Here is how the last year unfolded. The baseline. In May 2025, Google researcher Craig Gidney estimated that a quantum computer with fewer than one million noisy qubits could break RSA-2048 in under a week (arXiv). The previous estimate was 20 million, over a 95% decrease. Google cuts the cost of breaking blockchain signatures. In March 2026, Google Quantum AI published a new estimate for attacking the elliptic curve behind Bitcoin and Ethereum signatures. It was roughly ten times cheaper than the best previous estimate: about 1,200 to 1,450 logical qubits. Google judged the details too dangerous to publish. It withheld its circuits and released a zero-knowledge proof that they exist. AI-powered open challenge beats Google. At the end of May, ecdsa.fail turned Google’s work into an open challenge, where anyone could submit their own circuits and score against Google’s work. More than 100 people entered, most working with AI. They surpassed Google’s figure in under three days. Google’s figure came from one of the world’s top quantum research teams. AI powered collaborative open work by regular people sits 69.4% ahead of Google’s number. AI finds an attack NIST’s experts missed. In July 2026, Anthropic’s Claude Mythos Preview found a new attack on HAWK. HAWK had spent three years in NIST’s search for backup signature standards, surviving two rounds of cuts from a field of 40, and was the only lattice-based scheme to reach the final rounds. Years of expert review had missed the attack. HAWK was withdrawn the next day (Anthropic). AI starts doing new mathematics. In August, OpenAI said an internal model had solved 10 open problems, including new results on the lattice based mathematics that post-quantum cryptography is built on (The Quantum Insider). In October, OpenAI released 722 math papers written by AI. Many disproved assumptions mathematicians had relied on for decades (OpenAI, SiliconANGLE). Researchers sound the alarm. On October 7, Justin Drake, an Ethereum Foundation researcher and co-author of both the Google and ECDSA.fail papers, wrote that “mathematical superintelligence is upon us” (X). He stated it is reasonable to brace for ECDSA, the signature scheme behind Bitcoin and Ethereum, to break before Q-Day, “worst case in months not years.” – He called on the industry to plan for “bunker mode” (ForkLog). Then Vitalik Buterin named what came next. The same day, the Ethereum co-founder wrote that the “core new area of risk” from AI-driven math is ML-DSA and lattice-based cryptography, the same family of math as HAWK (X). Vitalik expects lattices’ real-world security may “take serious hits from the next two years of AI math.” His alternative is hash-based cryptography. ML-DSA is NIST’s main post-quantum signature standard; many teams today believe they are quantum resistant by using this scheme. The outcome of this? Regardless of quantum computing, the core cryptography meant to protect blockchains from quantum computers may not survive AI. What does this mean for blockchains? Most blockchains rest on a foundation of math, all of which face quantum and AI pressure. The list of “safe” algorithms is constantly in motion. NIST finalized ML-DSA in August 2024 (NIST). Two years later, it is being named as the next thing AI may weaken. HAWK was not the first NIST candidate to fall late. In 2022, two serious candidates broke on laptops: Rainbow, a signature finalist, fell in a weekend, and SIKE fell in about an hour (Quanta). NIST’s standards are still moving. In March 2025 NIST chose HQC, built on different math, as a backup to its lattice-based encryption standard, so that one weakness in lattices cannot take out both (NIST). It is also running a separate call for additional signature schemes, mainly ones not based on lattices (NIST). Post-quantum security is not a finish line. The algorithms called safe today are the best current guess, and AI is updating the guess faster than ever. What this all means in practice: Exposed keys are first in line. Any address whose public key is onchain is exposed, including every account that has signed a transaction. Busy signers carry the most risk. Oracles, bridge committees, security councils and validators sign constantly. Targets will move. A chain that hard-codes ML-DSA may need a second migration if lattices weaken. Migration can be slow, risky and expensive for any chain. What is secure today will not stay secure. So the question for any chain is not “which algorithm?” It is “how fast and safely can you change it?” What is cryptographic agility? Cryptographic agility is how quickly and safely a system can swap its cryptography without breaking or stranding its users. NIST describes it as the capacity to “seamlessly and rapidly transition away from vulnerable cryptographic algorithms and adopt new, stronger ones” (NIST CSWP 39). Quantum resistance solves the problems we can measure today. Cryptographic agility is how you handle the ones that come next: a quantum breakthrough, an AI-found attack, or a post-quantum scheme that turns out weaker than expected. How is Starknet built for cryptographic agility? StarkWare CEO Eli Ben-Sasson proposed what it takes for a chain to become future-proofed: the right cryptography, crypto agility, and control over its own upgrades. Starknet has the first two today and is exploring the final piece. Cryptographic agility: account abstraction. On Starknet, every account is a programmable smart contract, through a design called native account abstraction. Each account can switch to a new signature scheme and keep its address and funds, with no fork and no network-wide migration. OpenZeppelin has already demonstrated an account using Falcon-512, a post-quantum signature scheme, on Starknet mainnet. Falcon is lattice-based, like ML-DSA. If lattices become insecure, the same upgrade path can let an account move again. No Starknet account is locked into a single signature scheme. The right cryptography: hash-based proofs. Starknet’s proving layer is inherently quantum-resistant, as it is built on hash functions, not elliptic curve assumptions. StarkWare invented STARK proofs in 2018 and since then we have secured over one trillion dollars in cumulative trading volume using STARKs. This layer does not depend on lattices, so if AI erodes them, it does not need replacing. StarkWare CEO Eli Ben-Sasson compares quantum computers and AI to a tsunami, with each type of cryptography standing on higher or lower ground: Hash-based cryptography is the tallest mountain, 1,000 miles away from the ocean… It could be that all other cryptography is broken, and you still have hash functions. For the quantum tsunami, pick a really high mountain. For the AI tsunami, pick nothing but the highest mountain. Control over its own upgrades: becoming an L1. Today, Starknet depends on Ethereum for parts of its security, so it cannot fully secure itself on its own timeline. We are actively researching transitioning Starknet to an L1, so it can upgrade its own cryptography end to end and become the first fully quantum-resistant network, with 2027 as our target. The aim is to reduce inherited L1 exposure by bringing more critical security assumptions under Starknet-controlled infrastructure. Future-proof means cryptographic agility If you want a future-proof blockchain, you need cryptographic agility. Nobody knows which algorithm survives the next five years of AI math or quantum hardware. Starknet is built so it doesn’t have to know: hash-based proofs, accounts that can change their own cryptography, and a path toward controlling its own upgrades. Quantum resistance is how you survive now. Cryptographic agility is how you stay surviving. The next steps for us, from hash migrations to post-quantum consensus signing, are set out in our quantum hub.(Consensus expectations: Consensus for 2026Q3 expects EPS of 3.003 and revenue of USD 127.833 billion.)

25/100·CNeutral+12

news11/3news

Onchain finance has mostly offered two options: a glass house where every balance and transfer is public, or full anonymity that many regulated businesses won't use.

Aptos offers a third path: institutional privacy, through a confidentiality stack designed for compliant use cases.

It's opt-in, and addresses stay visible onchain, keeping a clear pathway to accountability. Three layers are live on mainnet.

1) Identity: Keyless Accounts let users create and manage blockchain accounts with existing credentials like Google or Apple, without publicly tying that identity to an onchain address.

2) Amounts: Confidential Assets is a standard designed for encrypted balances and transfer amounts, live today with APT. A zero-knowledge proof confirms each transfer is valid, while the amounts stay encrypted and sender and recipient addresses stay public.

3) Intent: Encrypted Mempool lets users opt in to encrypt a transaction's details until its block is finalized, designed to lower the risk of MEV attacks.

What this enables:

• Users can get started with Keyless Accounts using a login they already have, and use Confidential APT for payments where amounts stay between sender and receiver.
• Traders can encrypt the details of the orders they choose with Encrypted Mempool, with order details decrypted only after the block is finalized. AI agents' trade inputs can stay encrypted until execution, too.
• Institutions can encrypt balances and transfer amounts with Confidential APT, and place large orders without signaling intent before execution with Encrypted Mempool, while counterparty addresses remain visible onchain.

Why Aptos:

• Confidentiality is built into the protocol layer, not added as third-party tooling.
• Transparent by default, confidential by choice.
• The same network runs payments, trading, and settlement, with sub-second finality and parallel execution underneath.

Confidential APT is live, available via @PetraWallet and on Confidential Assets Web by @AptosLabs. Encrypted Mempool is available via Petra and on @DecibelTrade, and its transactions incur an additional gas fee.

• Petra: https://petra.app
• Confidential Assets Web: https://confidential.aptoslabs.com

62/100·B+Long+5

meme11/2meme

🇺🇸🇮🇷 Iran's FM @Araghchi on the Iran-US talks: The discussions have become very pressured. We have given put forward our conditions, our proposal. It is now in the US' hands, and how much political wisdom they have. Are they ready to take the path of diplomacy, or will they again go the military route. We are ready for both. We are more ready this time, for both. If they try to test us again, the result will be worse than last time. We have our conditions for the reopening of Hormuz. If they accept them, the strait will open and negotiations will resume. If not, then the situation stays as is. Retweeted by @staunovo from @ncitayim

72/100·B+Long

news11/2news

GoPlus October 2026 Web3 & AI Security Data Report Throughout October, Web3 recorded 45 major security incidents, with total losses of approximately $6,469,246,091 ($6.469 billion)—8.2 times September’s approximately $793 million and 34.4 times August’s approximately $188 million. October became the worst month for losses so far this year. The largest single incident, involving Binance’s multichain hot wallets, accounted for $6.3 billion. The top five incidents totaled approximately $6.423 billion, or 99.3% of total losses. Losses were overwhelmingly concentrated in exploits: 39 incidents accounted for approximately $6,465,088,419. Compared with September, October’s risk became even more concentrated in centralized operational risks and supply-chain risks. Three categories—compromised platform signing and permission systems (approximately $6.313 billion), supply-chain and distribution-channel attacks (approximately $90.13 million), and hot-wallet, private-key, and weak-randomness issues (approximately $29.77 million)—accounted for approximately 99.4% of total losses. Binance and Ledger alone accounted for approximately $6.387 billion, or 98.7% of the monthly total. This month’s figures reflect reported incident losses before deducting frozen or subsequently recovered assets. In AI security, October marked a shift from asking “Is the agent itself secure?” to examining “How does trust propagate across agents, protocols, identities, and tools?” AI-driven attack and defense also began operating at machine speed. AI agents carried out end-to-end real-world intrusions, while malicious instructions spread along internal trust chains in MCP/A2A environments. Hardcoding issues in IBM MCP Context Forge again exposed weaknesses in agent gateways’ default security. Both attackers and defenders are entering the machine-speed era. I. Web3 Security Overview 1. Overall Figures: October 2026 Incidents: 45 Total losses: $6,469,246,091 (approximately $6.469 billion) Exploits: 39 incidents, approximately $6,465,088,419 Phishing involving substantial losses: 6 incidents, approximately $4,157,672 Largest single loss: $6.3 billion (Binance) Top five incidents combined: Approximately $6.423 billion, or 99.3% of total losses Incidents with losses exceeding $1 million: 19 October’s incident count rose approximately 15% from September’s 39, while total losses increased to 8.2 times September’s level. The top five incidents’ share of losses rose from 94.8% to 99.3%. Four incidents exceeded $9 million in individual losses; two exceeded $80 million, and one reached $6.3 billion. Binance alone accounted for approximately 97.4% of the month’s reported losses, making it the decisive factor behind the surge. Meanwhile, the 15 incidents explicitly classified as “smart contract vulnerabilities” averaged approximately $590,000 in losses each. The pattern of frequent incidents with relatively low individual losses remained unchanged. 2. Major Attack Types October’s major attack types, grouped by losses, were as follows: Viewed by attack surface, five structural trends warrant particular attention: Trust roots and permission systems: Three incidents involving platform signing and permission systems caused approximately $6.3125 billion in combined losses. Binance’s central issue points to abuse of its signing or withdrawal systems, rather than a straightforward onchain contract vulnerability. At 79thVault, compromise of a privileged administrator address caused approximately $12.5 million in losses. The absence of direct private-key exposure does not establish that withdrawal and signing workflows are secure. Supply chains and distribution channels: Five incidents caused approximately $90.13 million in combined losses, spanning five entry points: a hardware-wallet reseller, a third-party RUM script, an API documentation website, wallet-extension source code, and a domain registrar. Attackers are looking beyond projects’ own code, prioritizing upstream nodes where one compromise can spread through multiple distribution paths. Account abstraction and signing mechanisms: EIP-7702 featured in the Binance platform-level incident, an individual’s approximately $3.55 million loss through malicious delegation, the AlpsCon[.]ai attack, and a vulnerability in Argent-style smart wallets. The Safe incident exposed another risk: granting token approvals to an address before any code has been deployed there. Weaponized governance: x100 lost approximately $4.85 million to a flash-loan governance takeover. Following BarnBridge/BonkDAO in July, term_labs in August, and Neutron in September, governance attacks appeared among major monthly incidents for the fourth consecutive month. Proposals have become a persistent attack vector. The long tail of contract and cross-chain logic flaws: Contract and business-logic vulnerabilities accounted for 19 incidents, with widely varying losses. Initia, Autheo, and Cysic again showed that flaws in VM execution security, business-logic implementation, or cross-chain minting permissions can cause severe financial losses. 3. Notable Incidents Binance: Attack on Multichain Hot-Wallet Signing Systems, Approximately $6.3 Billion in Reported Losses On October 30, Binance’s multichain hot-wallet infrastructure suffered large-scale unauthorized outflows, with reported losses of approximately $6.3 billion—the largest single security incident so far this year. Approximately $3 billion in stolen funds has been frozen. Binance.US operates independently and was unaffected. BNB Chain itself was not attacked, and the incident did not involve a protocol-level vulnerability in Ethereum or EIP-7702. Onchain evidence shows that the attacker obtained a valid EIP-7702 authorization signature for the Binance 14 hot wallet, delegated the wallet to a malicious contract, and then transferred ETH and 63 different tokens in batches. The attacker also gained control over USDC in Hot Wallet 34. EIP-7702 primarily increased the efficiency of asset transfers in this attack. Authorization signatures covering more than 50 consecutive nonces, coordinated theft across multiple chains, and directly signed transfers from other wallets make compromised signing infrastructure a more plausible explanation than the exposure of a single raw private key. The root cause, however, remains subject to the final investigation. The incident again demonstrates that a valid authorization signature does not establish that the signing action was genuine or legitimate. Ledger Reseller Incident: Tampered Hardware Wallets Cause Approximately $87 Million in Losses On October 10, an incident involving Ledger reseller CryptoBillis’s distribution channel led to widespread user losses. Some hardware wallets purchased through this channel had been tampered with before delivery, and assets subsequently deposited were stolen. The incident database records approximately $87 million in losses. A hardware wallet’s security boundary extends beyond its chip and firmware to production, logistics, distribution, activation, and authenticity verification. Confirming that a device works normally does not prove that it remains trustworthy. CoinDCX: Suspected Hot-Wallet Theft, Approximately $16 Million Lost On October 22, CoinDCX suffered a suspected hot-wallet theft involving approximately 15.99 million USDT and 5,783.81 TRX, with total losses of approximately $16 million. Following Bitget and Duelbits in September, hot wallets at exchanges and platforms holding large balances remained priority targets. For these platforms, the security of signing services, automated withdrawals, and backend operations systems often directly determines the losses and damage that can follow a single point of compromise. 79thVault: Privileged Administrator Address Compromised, Approximately $12.5 Million Lost On October 8, BNB Chain project 79thVault lost approximately $12.5 million through either a leaked private key for a privileged administrator address or insider activity. Nearly 10,000 token holders were affected. Administrator privileges remain one of the most dangerous single points of failure across many projects. Safe Multisig Address: Approvals Granted Before Code Deployment, Approximately $7.45 Million Lost On October 27, a vault associated with a Safe multisig address on Avalanche was exploited for approximately $7.45 million. The Safe had authorized an address to spend four tokens without ensuring that only trusted application code could execute at that address. No contract had been deployed at the address when the approvals were granted. An external actor subsequently used public deployment tooling to deploy their own withdrawal code there, then used the existing token approvals to transfer funds without another Safe signature. The incident shows that, in account-abstraction environments, trusting an address does not mean that code deployed there in the future can also be trusted. x100: Flash-Loan Governance Takeover, Approximately $4.85 Million Lost On October 24, x100 suffered a flash-loan governance takeover. Within a single transaction, the attacker acquired two of the Senate’s three seats and immediately executed a proposal to sell the treasury’s X tokens into a thin-liquidity pool they had just created, ultimately obtaining approximately 4.848 million USDT. Acquiring governance power, passing the proposal, executing it, and transferring assets all occurred within the same transaction. Governance defenses that rely on voting periods and manual review face substantial risk; safeguards such as timelocks are essential. Polymarket: Third-Party Monitoring Script Compromised, Approximately $2.71 Million Lost On October 29 (UTC), users of Polymarket’s international site suffered asset theft. A supply-chain compromise of third-party performance monitoring provider DebugBear’s RUM script delivered malicious code to user pages. The code abused users’ trading and signing permissions. Onchain analysis shows that the primary attack path involved 710 DepositWallets and 731 addresses, with approximately $2.71 million in losses. Address counts do not represent unique users. Every third-party script introduced into a frontend may gain access to user sessions and signing interactions. Software supply-chain security reviews must account for that exposure. Initia: MoveVM Type Confusion Enables Unbacked Minting, Approximately $2.73 Million Lost On October 22, a vulnerability in Initia’s MoveVM was exploited. The attacker induced a failure in the verified-script cache, causing cross-module ABI type confusion and bypassing normal MintRef handling and supply accounting to mint unbacked iUSD, AUSD0, USDC, and WETH. The assets were then moved cross-chain through LayerZero, IBC/Noble, and CCTP. Losses totaled approximately $2.73 million. 4. Security Recommendations Continuously review trading-platform signing and withdrawal workflows through an independent chain of trust. Validate transaction-data integrity, caller identity, and authorization context. Implement real-time circuit breakers for consecutive nonces, anomalous batch authorizations, and synchronized withdrawals across chains, alongside per-wallet limits and execution delays. Monitor EIP-7702 delegations and smart-account permissions explicitly. Clearly display delegation-target code and permission scope, and provide one-click revocation and anomaly alerts. Do not grant asset permissions in advance to addresses whose code is not yet determined or can be deployed by someone else. Extend supply-chain defenses across delivery and runtime. Purchase hardware wallets only through official channels and verify authenticity. Minimize third-party scripts, enforce integrity checks and CSP restrictions, enable registrar locks and multifactor authentication for DNS management, and establish reproducible builds and signed releases for wallet extensions. Minimize and isolate administrator and governance privileges, and enforce timelocks. Asset control, parameter changes, additional issuance, and upgrades must not be concentrated in one address. Enforce delays between acquiring governance power, executing proposals, and transferring treasury assets that cannot be bypassed within a single transaction. Generate wallet seeds and keys with cryptographically secure random number generators. Do not use randomness sources whose outputs can be feasibly enumerated, such as Math.random(). Assess wallets historically generated with weak randomness and migrate their assets. Establish continuous smart contract auditing to prevent financial losses caused by contract vulnerabilities. II. AI Security Incidents and Trends If September’s AI security themes were “data assets and authorization boundaries,” October made another shift clear: authorization is flowing across agents, protocols, and identity systems, while AI agents are beginning to execute attacks end to end at machine speed. The month’s notable incidents and developments reflect changes across real-world attack surfaces, protocol trust, endpoint rendering, gateway control planes, and frontier offensive and defensive capabilities. AI-Driven Multicloud Intrusions: Agents Are No Longer Just Assisting Attackers On October 19, Wiz CIRT published findings from investigations into multiple real-world intrusions. AI agent systems were no longer merely advising attackers. They were executing entire workflows: finding entry points, collecting credentials, moving laterally across platforms, exfiltrating data, and debugging scripts. One representative case began with an exposed Kafka UI Java remote code execution vulnerability, then spread across Kubernetes, GCP, AWS, CI/CD, GitOps, and identity providers. The agent generated cloud credentials in bulk, obtained cluster-admin privileges, created backdoor identities with administrator access, and ran five workflows simultaneously: secret collection, reachability mapping, credential spraying, service exploitation, and compute expansion. These attacks performed hundreds to thousands of actions per hour. After failures, the agents read error messages, rewrote scripts on the fly, and retried. The same source IP also scanned more than 600 other Wiz customer environments during the same window. The activity was not stealthy—it was conspicuously noisy. But its traces were scattered across cloud control-plane, identity, CI/CD, and workload logs, leaving no single alert capable of reconstructing the full attack. The defining change is not a more sophisticated individual exploit. It is lower execution costs for cross-platform attack chains and trial-and-error cycles compressed to machine speed. MCP/A2A “Protocol Hopping”: Malicious Instructions Propagate Along Internal Trust Chains On October 6, security researchers disclosed how attackers could first manipulate an internal agent, then pass malicious instructions to other agents as routine delegated tasks. Downstream agents trusted upstream agents by default, while MCP servers held credentials for each agent. Prompts that model guardrails might otherwise reject could trigger data access, outbound requests, and even SSRF after being forwarded across protocols and undergoing identity transformations. The research covered MCP, Google A2A, and Agent Network Protocol, among others. In Google MCP Toolbox for Databases, the HTTP client lacked redirect checks and destination-IP validation. Attackers could craft path parameters that caused the tool to follow redirects to internal endpoints. The issue was fixed through network-range allowlists and blocklists, alongside startup validation. The core issue in this kind of protocol hopping is a failure in how authorization and context are handled across protocols. Internal agent-to-agent calls can no longer be treated as inherently trusted internal-network traffic. Identity, permissions, and task scope must be revalidated. IBM MCP Context Forge Hardcoded JWT [REDACTED CREDENTIAL] Admin Tokens Can Enable Gateway Takeover In October, CVE-2026-53709, a hardcoded JWT secret vulnerability in IBM MCP Context Forge, drew attention with a CVSS score of 9.8. Affected versions—1.0.1 and earlier—shipped the same default JWT_SECRET_KEY in the source code and .env.example, while defaulting to REQUIRE_USER_IN_DB=false. Anyone with access to the public repository could forge platform-administrator JWTs without an account, password, or valid user record in the database. Deleting administrator accounts, changing passwords, and revoking sessions could not stop an attacker from generating fresh valid tokens. Platform administrators could read and modify registered tools, gateways, servers, prompts, and resources, access some credentials, and register malicious upstream MCP servers. A gateway compromise could thus become a supply-chain attack against every connected client. The issue was fixed in version 1.0.2 and later. Agent gateways are becoming a new control plane. Default keys, default administrators, and permissive initialization settings can be as dangerous as default passwords on traditional network appliances. AI Security Recommendations Enforce zero trust at every agent-to-agent hop. Revalidate identity, audience, permissions, and task scope whenever a request crosses MCP/A2A or another protocol. Do not inherit implicit trust from upstream agents. Apply destination allowlists and redirect validation to tool-generated outbound traffic. Build threat intelligence and automated response for attacks operating at machine speed. Every team needs an always-on AI CISO—Chief Information Security Officer—to continuously track threats, assess business impact, coordinate risk validation, and develop actionable remediation and incident-response plans. Configure agent gateways to fail closed. Block startup with default or weak signing keys. Automatically generate a unique, strong key on initial deployment, and bootstrap administrator access using one-time tokens and actual user identities recorded in the database. Upgrade MCP Context Forge promptly, and audit gateway registrations and upstream tool changes. III. Summary and Outlook In October 2026, Web3’s approximately $6.469 billion in reported losses and the top five incidents’ 99.3% share were driven sharply higher by the exceptional Binance incident. Excluding Binance, losses totaled approximately $169 million, down about 78.7% from September. Yet 45 incidents, 19 with losses exceeding $1 million, and a fourth consecutive month of governance attacks show that the risk surface has not narrowed. Defense must expand beyond private-key protection and contract auditing to include signing-context validation, supply-chain integrity, and governance of account execution environments. In AI security, real-world end-to-end agent intrusions, cross-protocol propagation of malicious prompts, and default gateway keys show that guardrails at the model-output layer are insufficient. Identity, permissions, tools, rendering, and distribution must all be brought under engineering controls. Frontier models discovering and exploiting complex vulnerability chains at scale also require defenders to adopt AI for continuous scanning, rapid validation, and automated response. For Web3 and AI, the next stage of security must extend beyond patching individual vulnerabilities to governing trust roots, enforcing authorization at every hop, validating the full supply chain, developing AI-driven offensive and defensive capabilities, and coordinating defense at machine speed.(Historical earnings: For fiscal 2026Q3 (period ended 2026-09-30), IBM reported basic EPS of 5.92, diluted EPS of 5.85, and net income of USD 5.571 billion. Consensus expectations: For the upcoming 2026Q4 event on 2027-01-26, consensus EPS is 4.6122 and consensus revenue is USD 20.468 billion.)

78/100·AShort+23

meme11/1meme

"The point of a debate isn't to win. It's to find the best path forward." I heard that on a Zcon panel this past week. It's more true now than ever. I still have concerns with this proposal, mainly how the Future Dev and Legal Defense Fund lockbox is ultimately controlled and disbursed. But in the interest of moving forward and bringing private money to billions, I see this as a reasonable compromise. Onward 🛡️ Retweeted by @zooko from @tonymargarit

10/100·CNeutral

meme11/1meme

10 Spicy Takes After Talking to 50 Friends at the Frontier of Robotics 🤖🔥 1/ Astra IS robotics' GPT-3 moment. Ironically, now that it's finally here, some in robotics seem reluctant to acknowledge it. Perhaps because it (sadly) didn't come from a robotics company? 2/ Robotics is becoming just another vertical for frontier labs. Not so different from coding in some respects. Why reinvent general intelligence for every industry? 3/ Robotics was too hard to underwrite until now. No convincing pretraining scaling law, no convergence on models, data, or hardware. No wonder some critics dismissed it as a fragmented industry playing an expensive "demo game". Astra may finally offer a credible path to scale. 4/ Robotics foundation-model startups have a case. But how strong is it? The argument: Astra handles System 2 (reasoning), while specialized models retain System 1 (skills) and System 0 (motor control). Fair. But isn't that already a narrower value proposition? And who says the boundary stops there? 5/ Follow the talent. Word is OpenAI reorganized its 30–40-person robotics team around the core Astra model. Hearing similar moves toward general models at @Tesla_Optimus. Org charts may tell us more than any demo. 6/ The robotics data gold rush is real. Million-hour procurement programs, some potentially reaching tens of millions of hours. Talk of budgets approaching $1B per frontier lab. 7/ Bullish on robotics hardware volumes. As intelligence becomes more accessible, the barriers to building useful robots could fall dramatically. The moat shifts toward reliability, dexterity, integration, and deployment. 8/ Robotics could fuel the next semiconductor and memory supercycle. A 100B-parameter model running on the edge requires ~50GB just for 4-bit weights. Now multiply that by millions of robots. Why do you think @JensenHuang is so obsessed with robotics? @elonmusk saw it coming: https://x.com/elonmusk/status/2114772074438988256?a=ZJpCuCM8Iv8RkRR_g_A&s=20 9/ Is self-driving immune to the general-model threat? After coding and robotics, why assume self-driving remains the exclusive domain of proprietary, vertical models? Robotaxi increasingly feels like a solved problem to me. But if and when frontier labs move into self-driving, we could see an Apple vs. Android dynamic. Think about the implications for $TSLA, Waymo, and other OEMs. 10/ The next bottleneck may not be intelligence. It's deployment. Models are improving faster than our ability to deploy robots in the messy real world. Integration, hardware variance, failure recovery, and economics remain brutally hard. The real moat may be turning intelligence into reliable, profitable physical work. -- I realize some of these takes may rub people the wrong way, especially those who have spent years doing the hard work of advancing robotics. That's not my intention. I'm sharing them to spark debate, not to declare winners and losers. Astra has only been out for a month or so, and it's far too early to draw definitive conclusions. More importantly, when the TAM is this enormous - potentially the largest in tech - there should be plenty of room for different approaches and many winners. ++ Full diligence report: https://robonomics.substack.com/p/robotics-new-era-after-astra?r=3hcp4&utm_campaign=post&utm_medium=web Retweeted by @jukan05 from @FredaDuan(Historical earnings: For 2026Q3 (period ended 2026-09-30), Tesla reported basic EPS of 0.74, diluted EPS of 0.67, net income of USD 2.378 billion, and revenue of USD 78.669 billion. Consensus expectations: For 2026Q4, consensus EPS estimate is 0.4596 and revenue estimate is USD 29.489 billion.)

10/100·CNeutral+14

meme10/31meme

#Iran’s SNSC Secretary: No new decision has been made on changing our #nuclear doctrine, though keeping the old doctrine if war resumes is under discussion. The Iranian nation has demands; seeing the enemy attack nuclear sites, people naturally ask why continue the previous path. Retweeted by @staunovo from @IranNuances

72/100·B+Long

Twitter10/31news

For the record, I don't support the grant and am not defending it. I believe that Dev Funds should be as apolitical as possible. I'm not involved with ZCG or their decision-making.

It's possible to engage thoughtfully with ZCG and reason without character assassinations.

I maintain that the best path forward is to privately fund everything and close the Zcash Dev Fund entirely. Free markets.

25/100·CNeutral

news10/30news

Crypto's Biggest-Ever Hack: How $6.3 Billion Was Stolen From Binance On October 30, Binance @binance suffered an estimated $6.3 billion hack, the largest crypto theft on record. Attackers drained hot wallets across multiple blockchains in under four hours.Binance suspects its internal systems were compromised but found no evidence of leaked private keys. The full attack path remains under investigation. Attackers then converted some stolen USDT into ETH and TRX, while depositing 60 million USDT into Aave to borrow 19,762 WETH ($44.6 million), shifting funds into assets that cannot be directly frozen by token issuers.

88/100·AShort+3

news10/29news

🚨NEW: A Nasdaq spokesperson declined to comment on claims that its representatives urged listed companies not to have their shares tokenized through OKXICE. The spokesperson reiterated that Nasdaq plans to stay out of venues operating under the SEC’s innovation exemption, saying it does not view them as the best path for tokenization.

“We remain committed to tokenization and are actively building that future through the Nasdaq Equity Token, an issuer-centric model that brings the benefits of tokenization into the established market framework.”

The exchange again denied involvement in this week’s WSJ ad, saying it had not authorized the ad to use or link to its FAQ, including through its QR code.

25/100·CNeutral+2

news10/29news

Strong US inflation and growth will keep Fed on a tightening path

72/100·B+Short

meme10/29meme

@chainlink @The_DTCC Production trades in July and a live service in October is a fast path for market infrastructure of this size. Congrats to @The_DTCC, and to @chainlink for the work behind it. Reply to @chainlink: Congratulations to @The_DTCC on the launch of the DTCC Tokenization Service, a significant milestone in the modernization of global financial market infrastructure. The launch follows DTCC’s first-ever production trades of tokenized U.S. securities in July 2026, powered by Chainlink alongside more than 30 major institutions: • BlackRock • J.P. Morgan • Goldman Sachs • Vanguard • NYSE • Nasdaq • CME Group • Microsoft • State Street Investment Management • BNP Paribas Securities Corporation • Societe Generale • S&P Dow Jones Indices • Citadel Securities • FTSE Russell • Invesco • Broadridge • Tradeweb • + many more With quadrillions processed annually as critical infrastructure underpinning U.S. securities markets, DTCC is bringing tokenization to the American financial system. We look forward to continuing our work together in advancing 24/7 collateral workflows through the Chainlink Runtime Environment (CRE) & data standard, accelerating institutional adoption of tokenized assets.

72/100·B+Long+5

meme10/29meme

Federal Reserve Chairman Kevin Warsh on Wednesday left the central bank on track to raise interest rates again this year, even as he stuck to his refusal to offer any explicit guidance on the future path of policy https://www.bloomberg.com/news/articles/2026-10-28/fed-holds-fire-as-warsh-telegraphs-more-hikes-without-saying-it?utm_source=website&utm_medium=share&utm_campaign=twitter

72/100·B+Short

meme10/28meme

Co-validation is live on Canton Network. Ledger Wallet now supports $CC, with transactions requiring confirmation from both a Ledger-operated and a Kiln-operated validator before Canton accepts the resulting state. Read more: https://kiln.fi/post/ledger-wallettm-x-canton-a-co-validation-model-operated-with-kiln https://x.com/Kiln_finance/status/2114660072345231727 Quoted @Kiln_finance: Ledger Wallet™ now supports @CantonNetwork 🟢 Kiln operates the second validator in a co-validation model with @Ledger. Both validators must confirm every transaction. Keys came from an air-gapped ceremony, and Ledger’s recovery path is set out in the contract. How it works 👇 https://www.kiln.fi/post/ledger-wallettm-x-canton-a-co-validation-model-operated-with-kiln

72/100·B+Long

news10/28news

AI Crisis: Smart Contract Bug Hunting & Breaking Encryption Executive Summary AI’s impact on digital asset security has moved beyond generating audit assessments to discovering vulnerabilities, constructing attacks, and validating their outcomes. Smart contracts are particularly exposed to this shift: contract code and historical transactions are typically public, execution results can be reproduced, and financial losses can be measured directly through state changes. Attackers need only one viable path to succeed. Defenders must identify and close every gap. Traditional, one-off audits struggle to address this asymmetry. Developments in cryptanalysis also demand a reassessment. Describing AI as capable of “only finding implementation bugs” no longer accounts for the research published in 2026. New structural key-recovery attacks have emerged against the HAWK signature scheme, alongside new algebraic methods and improvements in attack complexity against seven-round AES. However, the HAWK results do not break lattice-based cryptography as a whole, and the seven-round AES results do not break full-round AES. [5–9] This review reaches six key conclusions: Automated smart contract exploitation now has empirical support. SCONE and EVMbench demonstrate AI’s ability to validate vulnerabilities that cause financial losses through end-to-end exploitation in controlled environments. Benchmark success rates, however, cannot be read as the proportion of all mainnet contracts that could be successfully attacked. The AI arms race in smart contract security is underway, and defenders still have an opportunity to respond. [1–4] AI-assisted original cryptanalysis now has cases that warrant serious attention. HAWK, reduced-round AES, and certain design flaws examined in CryptanalysisBench show that AI is beginning to make substantive contributions to the analysis and breaking of cryptographic algorithms. These developments demand close attention and internal reviews of potentially affected systems. [5–9] Mainstream cryptographic systems have not been broken as a whole. No research has yet demonstrated that AI can break correctly implemented, appropriately parameterized full-round AES, RSA-2048, secp256k1, or standard ML-KEM. The most immediate losses remain concentrated in contract logic flaws, poor private-key custody, phishing, and randomness vulnerabilities. Exploiting these weaknesses generally does not require solving the underlying mathematical problems. AI’s research speed will amplify gaps in validation capacity. Producing findings is not the same as identifying valid vulnerabilities, much less managing the associated risks. Vulnerability discoveries, theoretical attacks, and patches still require human review and validation. This poses a major challenge for security teams. The security paradigm needs to evolve quickly. Integrate AI into continuous auditing, runtime monitoring, and response. Keep weak points—such as private-key management and fund-management permissions—under effective controls outside the AI system. Build an inventory of cryptographic assets and the ability to migrate algorithms as soon as possible. I. Why Smart Contracts Are Particularly Exposed to AI-Driven Vulnerability Discovery 1.1 Observable, Reproducible, with Quantifiable Losses The openness of smart contracts makes them well suited to AI-driven vulnerability discovery. AI can typically access contract code and state without restriction, formulate testable hypotheses, reconstruct execution conditions locally, and evaluate outcomes through transaction traces and balance changes. This environment provides clear feedback and lends itself to automated iteration. Openness does not make the process cost-free. Contracts without publicly available source code, proxy implementations, compiler settings, external protocol state, onchain time, and asset liquidity can all determine whether an attack is viable. AI’s advantage lies in making these conditions faster to understand and explore. 1.2 Contract Security Depends on Code, Economics, and Permissions A system can contain vulnerabilities in asset pricing, cross-contract state, or governance workflows even when no individual function has an obvious coding error. The price used to value collateral may be temporarily distorted. A router trusted by multiple vaults may fail to enforce distinct permissions for each vault. A legitimate upgrade may introduce an incompatible storage layout. Investigating these issues requires understanding three things at once: what the code permits, what the application is intended to permit, and what an attacker can control. Advances in AI’s ability to search large codebases and explore combinations of hypotheses make investigations that once demanded extensive manual effort easier to scale. 1.3 Lower Discovery Costs Expand the Pool of Economically Viable Targets As analysis costs fall, smaller protocols, forks of older versions, and long-unmaintained contracts may become economically viable targets. Repeated deployments of the same underlying code amplify the impact: a single vulnerability may affect instances across multiple projects. These economies of scale make AI-driven vulnerability discovery a highly cost-effective approach for attackers in the smart contract ecosystem. II. Smart Contract Vulnerability Discovery: What Public Research Has Demonstrated 2.1 SCONE: Retesting Historical Exploits In the SCONE study published in 2025, Anthropic and collaborating researchers used 405 contract tasks drawn from real attacks between 2020 and 2025, covering Ethereum, BNB Smart Chain, and Base. All experiments ran in simulators or on local chain forks. [1] The study evaluated 10 models. Across all models and attempts, 207 tasks were solved—51.11% of the total—with aggregate simulated profits of approximately $550.1 million. A May 2026 research update selected another 12 tasks based on historical attacks reported after January 1, 2026. The researchers reported that Mythos Preview found successful exploits for all 12, generating approximately $35 million in simulated profits. [2] 2.2 EVMbench: Finding, Patching, and Exploiting Are Different Tasks The arXiv v1 paper for EVMbench was published in March 2026 and included 117 high-severity vulnerabilities from 40 code repositories. This version evaluated vulnerability discovery, patching, and exploit execution separately, with different task counts across the three modes. [3] The researchers also found that providing business descriptions significantly improved results on some patching and exploitation tasks. The same model could also perform differently under different agent frameworks. These findings show that identifying problems in large codebases, using tools, and organizing tasks are all important variables in AI-driven vulnerability discovery and exploitation. 2.3 GoPlus AI Auditing Benchmark: From Real-Time Incident Tracking to Unknown Vulnerability Discovery Released by GoPlus in May 2026, the AI Auditing Benchmark is a dataset built specifically to evaluate AI-powered smart contract auditing. It is grounded entirely in real-world attacks, covering both major incidents with substantial impact and losses, and smaller incidents with limited losses that are technically representative. It curates the source code of vulnerable contracts from attacked projects to systematically assess AI’s capabilities in vulnerability identification, attack-path comprehension, contextual reasoning, and audit consistency. The project incorporates security incidents as they occur each week and is continuously updated. It preserves the complete source tree of the attacked contracts, including application contracts, dependencies, and interfaces. This makes it suitable for evaluating AI auditing across complex call relationships, cross-module state flows, and the full context. More details: https://github.com/GoPlusSecurity/ai-auditing-benchmark 2.4 What These Benchmark Results Tell Us Together, these datasets and results demonstrate AI’s ability to validate vulnerabilities that cause financial losses through end-to-end exploitation in controlled environments. Benchmark success rates, however, cannot be read as the proportion of all mainnet contracts that could be successfully attacked. The AI arms race in smart contract security is underway, and defenders still have an opportunity to respond. III. How AI Is Changing Smart Contract Vulnerability Research 3.1 High-Value Vulnerability Classes and Validation Priorities 3.2 Refine Business and Security Descriptions Instead of Relying on Broad Vulnerability Keywords AI’s most valuable role is to turn code, deployment information, and business descriptions into testable questions, then use the appropriate tools to validate them. The following closed-loop workflow is recommended: Model assets and permissions → Define security properties → Analyze candidate mechanisms → Perform independent testing and solving → Inspect transaction traces → Confirm impact through human review → Apply fixes and run regression checks. Useful business and security knowledge should capture the protocol’s own commitments and key risks. For example: Ordinary users must not be able to change the administrator, implementation address, or protocol fee recipient. A one-time authorization must not be successfully reused within the same authorization domain after it is first successfully consumed. With fixed valuations, no external yield, and an explicit rounding policy, a round-trip exchange must not create additional redeemable assets out of thin air. Reprocessing a cross-chain message that has already been successfully processed must not increase the spendable balance again. When the contract is paused, fund operations that the documentation specifies should be paused must actually be unreachable. Even a property such as “assets exceed liabilities” requires explicit valuation and market assumptions. Market risk inherent in the intended design must not be misreported as a software vulnerability. Formal tools can prove only the properties written into the specification; they cannot replace judgment about whether the specification itself is correct. IV. Breaking Cryptographic Algorithms: What Needs Reassessment in 2026 4.1 HAWK: A Structural Attack with Consequences for Scheme Design In July 2026, Anthropic published research on key recovery against HAWK, a post-quantum signature candidate based on the lattice isomorphism problem. The research exploits nontrivial automorphisms in a particular lattice structure to reduce key recovery to lower-dimensional instances of the exact Shortest Vector Problem (SVP). It does not exploit a conditional branch that a programmer implemented incorrectly. [5–6] The technical paper gives a more precise result than headlines claiming that “key strength has been halved”: key recovery for HAWK-n can be reduced to a polynomial number of calls to an exact SVP oracle in dimension n/2 + 1. A polynomial-time reduction does not make the complete attack polynomial-time; the SVP oracle itself remains computationally expensive. On July 29, 2026, the HAWK authors announced their withdrawal from NIST’s additional digital signature standardization process, explicitly stating that the scheme was insecure and should not be used. The announcement noted that fixes such as simply increasing the parameters would leave it uncompetitive. [7] This is an algorithm-design result with substantive consequences. Its scope is also clear: the failure of HAWK’s particular structure does not mean that LWE, ML-KEM, or all lattice-based signatures have been broken. The paper also explicitly analyzes why the attack does not transfer directly to Falcon. As for AI’s contribution, the researchers describe work conducted in an agent environment equipped with tools, supported by computational experiments and intermittent human guidance. The work took approximately 60 hours and incurred about $100,000 in API costs. The algorithmic result and the scheme’s withdrawal can be independently verified. Precise attribution of how much AI completed independently still depends on the researchers’ process records. 4.2 Reduced-Round AES: Major Complexity Improvements, Still Far from a Practical Break Another 2026 paper examines the seven-round version of AES-128. Standard AES-128 uses 10 rounds; AES-192 and AES-256 use 12 and 14 rounds, respectively. [8][22] The paper builds on a meet-in-the-middle attack approach, using the algebraic structure of the AES S-box to construct a new fingerprinting method. This removes the need to guess one key byte. Together with computational optimizations, the method reduces attack costs. The authors report that, at a data complexity of approximately 2^105 chosen plaintexts, time complexity falls from roughly 2^99 to 2^89.3–2^91.4—a speedup of about 200–800 times, depending on the cost-accounting method. [8] The result has research value: it shows that AI-assisted research can produce new algebraic attack constructions against a long-studied target. It does not, however, constitute a practical break of AES as deployed in production: The target is a seven-round variant. The attack cannot be applied directly to full, 10-round AES-128. The required number of chosen plaintexts is enormous. This is not an attack that becomes usable after intercepting a small amount of TLS traffic. A time cost on the order of 2^89 or higher remains impractical. The original paper explicitly states that the complete attack was not run end to end; its components were validated through mathematical analysis and smaller-scale experiments. A “hundreds-fold speedup” describes a complexity improvement within a particular attack model, not the speed of decrypting arbitrary real-world ciphertexts. The paper and accompanying research overview also expose a validation bottleneck: generating a candidate breakthrough may take less time than humans need to establish its correctness and full cost. For future cryptanalysis review, this has greater long-term significance than this particular speedup. 4.3 CryptanalysisBench: Distinguishing Design, Specification, and Code Flaws The July 2026 v2 paper for CryptanalysisBench includes 191 tasks spanning six classes of cryptographic primitives: block ciphers, hashes, authenticated encryption, KEMs, public-key encryption, and digital signatures. The tasks range from schemes with known practical attacks to schemes without known practical attacks and their reduced versions, extending to a more difficult challenge set. [9] The study uses an isolated security controller to hold secrets, allowing AI to interact only through permitted queries. Final scripts are rerun on fresh random instances. This provides stronger validation than asking a model to guess a key that is already public. The paper reports that five models achieved solve rates of approximately 65%–86% on Tier 1 schemes with known practical attacks. In Tier 2, results aggregated across models showed that 14 distinct schemes were broken in their full configurations. The authors further distinguish: Two schemes with genuine design or security-proof flaws: SpoC and KINDI. Two schemes with specification gaps that lead to insecure implementations. Ten schemes with errors in their reference implementations. For SpoC, the result concerns a design flaw in how the authenticated encryption mode handles boundary-case messages. For KINDI, it exposes an error in the scheme’s CCA security argument. These are substantive advances in the study of candidate schemes. They must not be presented as “AI breaking 14 mainstream standardized cryptographic algorithms.” The paper also explicitly acknowledges limitations: many attacks can be recalled from existing knowledge, reference-code errors affect overall scores, reduced parameters do not always produce a smooth difficulty gradient, and probabilistic security games introduce issues involving statistical errors and resource budgets. V. Why Insecure Implementations and Key Leakage Remain the Most Immediate Risks 5.1 Attackers Use Information Beyond the Ciphertext A cryptographic algorithm may be secure within its mathematical model, while its implementation on a device still leaks secrets through power consumption, electromagnetic emissions, timing, or cache behavior. ASCAD studies masked AES implementations and publishes measurement data and model configurations to improve the reproducibility of deep-learning side-channel research. [12] It demonstrates that mathematically correct AES implementations still need protection against leakage. It does not show that supplying a ciphertext remotely is enough to recover an arbitrary AES key. Evaluating this research requires explicit assumptions: Can the attacker access the target device? Can they obtain a comparable device whose keys they control? Are training and attacks performed on the same device? Does the attack transfer across devices? How many measurements are required, and under what noise conditions? Constant-time implementations primarily limit timing-related leakage; they do not replace dedicated defenses against power, electromagnetic, and other side channels. 5.2 Password Guessing: Learning Human Preferences, Not Eliminating Random-Key Entropy Research such as PassGAN learns human password distributions to improve candidate generation. [13] For some password distributions, a better ordering of guesses does increase the hit rate within a fixed budget. But password-hash verification, online login attempts, and searches for random cryptographic keys are three different problems. Offline password attacks are constrained by KDF costs. Online attacks also face rate limits and authentication policies. Learning common human passwords offers no equivalent advantage against independently and uniformly generated keys. Salts prevent direct reuse of the same precomputed results across accounts, but they do not make weak passwords strong. Memory-hard KDFs such as Argon2id help defenders by increasing the cost of each guess. They still need to be combined with sufficiently strong passwords, parameters appropriate to the device, and sound authentication design. [21] 5.3 Randomness and Nonces: Different Protocols, Different Requirements Reused, biased, or leaked ECDSA signing nonces can turn private-key recovery into an algebraic or lattice-analysis problem. AI can help identify errors, organize observations, and invoke analysis tools. The root cause of the failure is a violation of nonce requirements—not a solution to the general elliptic-curve discrete logarithm problem. RFC 6979 specifies deterministic nonce generation for DSA and ECDSA, an established mitigation for certain risks associated with dependence on randomness. [15] It does not mean that any custom deterministic nonce implementation is secure, nor does it eliminate side-channel attacks, fault injection, or poor key management. Nonces in AEAD have different semantics. For example, reusing a nonce under the same key in GCM can compromise confidentiality and authentication. Signing nonces, encryption nonces, and contract-authorization nonces must not be treated as the same mechanism. 5.4 Key Extraction, Unintended Signing, and Misuse If an agent can directly read seed phrases or invoke a signer without restrictions, attackers may influence its behavior through prompt injection, malicious dependencies, or tool outputs. Funds can still be transferred using valid signatures even when the underlying cryptography remains entirely sound. HSMs, MPC-based systems, and hardware wallets can reduce certain key-leakage risks. But if their policies allow arbitrary transactions to be signed blindly, they cannot fully prevent authorization abuse. The target contract, assets, amount limits, validity period, and transaction semantics must be validated outside the model. VI. Key Risks at the Intersection of Smart Contracts and Cryptographic Systems 6.1 Cryptographic Validity Does Not Guarantee Correct Business Authorization A signature can be mathematically valid without being bound to the chain, contract address, action, amount, recipient, or expiry. The verifier confirms that the signature corresponds to a particular key but does not establish whether that authorization applies to the current operation. EIP-712 provides structured signing and domain separation, but the specification explicitly states that it does not itself include replay protection. [14] One-time authorizations still require application logic to maintain nonces or usage records and verify that the fields covered by the signature exactly match the parameters used in execution. For bridges, proving that a message genuinely exists on a particular chain does not establish that the current contract should execute it. Requirements for message origin, destination domain, sender, uniqueness, and finality must all be satisfied. 6.2 A Correct Proof System Can Still Prove Too Little Successful verification of a zero-knowledge proof establishes only that a defined relation has been satisfied. If the circuit omits range constraints, fails to bind critical public inputs, or the verifier uses an incorrect verification key or verification context, application security can still fail. AI can help compare circuits, specifications, and onchain execution paths to identify gaps where “A was proved, but B was required.” Reports must specify whether the issue lies in circuit constraints, protocol binding, or the proof system itself. Such findings must not be indiscriminately presented as “zero-knowledge cryptography has been broken.” 6.3 AI Security Tools Introduce New Trust Boundaries Repositories under audit, dependency packages, error logs, and web content may contain malicious instructions targeting the agent. If that content influences its use of privileged tools, the research system itself may be induced to leak secrets, modify the validation environment, or invoke a live signing service. OpenAI’s documentation discusses prompt injection, untrusted input, data leakage, tool permissions, and structured-data boundaries, reflecting these general risks. [20] Specific protections should be enforced by the runtime environment and policies, rather than relying solely on prompts telling the model to be cautious. Keep the analysis environment, evidence-validation environment, and live signing environment separate. The analyzer can propose transactions or patches; an independent system checks their properties; the signer signs only as permitted by explicit policies. Sensitive actions must not be approved solely because “AI judged them safe.” VII. The Economics of Attack and Defense: How Falling Attack Costs Translate into Real Risk 7.1 Cost Model A simplified expression for the expected return from vulnerability discovery and exploitation is: The probabilities and values depend on protocol state, target selection, and resource budgets, and are generally interrelated. The expression provides a framework for analysis; it cannot predict theft amounts without supporting data. AI may reduce the cost of understanding code and constructing attacks. Better target selection may also reduce time spent analyzing unviable targets. Competition, slippage, and capital constraints, however, may offset some of those gains. Model invocation costs do not represent the full cost of an attack. 7.2 Attackers Need One Success; Defenders Bear the Burden of Coverage Attackers may need to find only one path that can be monetized. Defenders remain responsible for permissions, security specifications, deployment configurations, third-party dependencies, and emergency response. A single AI audit that finds no issues is no reason to stop checking. Defenders also have advantages: early access to code, design intent, deployment plans, and monitoring permissions. They can eliminate issues before launch and maintain continuous auditing and security monitoring afterward. Turning that information and time advantage into an ongoing validation process matters more than purchasing any particular model. 7.3 Validation and Remediation May Become Bottlenecks As AI increases the volume of candidate findings, security teams may be overwhelmed by low-quality reports unless they have deduplication, evidence grading, and counterexample reproduction in place. In cryptographic research, cost estimates and the scope of theoretical results also require more rigorous specialist review. Improving defensive capability requires speeding up all three stages: finding problems, confirming them, and fixing them safely. Simply counting reports, lines of code scanned, or model calls can obscure the real risks. VIII. Security Hardening Roadmap 8.1 Move from One-Off Audits to Continuous Auditing Project teams and developers should use specialized AI auditing tools for regular and continuous smart contract audits. These audits should provide greater depth and efficiency than traditional manual audits while remaining cost-effective. AI should analyze contract source code, permission management, and business logic in depth, systematically scan for common security issues—including syntax-level and business-logic vulnerabilities—and produce structured audit reports. Projects already in production should adopt continuous security monitoring and risk-response services. These services should closely track emerging vulnerability intelligence, newly reported attacks, and risks involving external dependencies, including vulnerable contracts, oracles, cross-chain bridges, and third-party protocols. AI should continuously assess whether the project is affected and provide targeted security-check plugins. This establishes ongoing security monitoring and risk response, addressing the limitations of traditional audits that treat the audit itself as the finish line. More details: https://deepscan.gopluslabs.io/ 8.2 Cryptographic Engineering: Close Directly Exploitable Gaps First Use established cryptographic libraries and standard cryptographic constructions. Avoid inventing custom combinations of encryption and signatures. Verify the correctness of randomness sources, nonce lifecycles, error handling, domain separation, and key derivation. For hardware and embedded deployments, assess the actual leakage channels. Build an inventory of cryptographic assets that records, at a minimum, algorithms, parameters, protocol uses, key locations, responsible owners, required data-confidentiality lifetimes, and replacement difficulty. Plan post-quantum migration for data that must remain confidential over the long term. For hybrid schemes, validate the composition, negotiation binding, and downgrade protection. Simply invoking two algorithms side by side is not enough. For schemes such as HAWK that their authors have explicitly declared insecure, review actual deployments and act according to the published advisory. For standard algorithms without evidence of an applicable attack, do not rush to replace them with untested designs based solely on headlines. 8.3 Agentic Security: Strict Controls over Funds and Private Data By default, AI analysis environments should hold no real seed phrases or production keys and should have no access to unrestricted signing interfaces. Scope AI tool access to the task. Restrict network egress and access to sensitive directories and blockchain endpoints. Treat untrusted code and logs strictly as data. Text within them must not change tool permissions. Validate the target, amount, asset, expiry, and permitted call semantics of every signing request. Set limits on amounts, frequency, and cumulative losses, along with stop conditions, for automated actions involving funds. Retain necessary records of inputs, tool calls, and validation while keeping secrets out of logs. Human review should focus on decisions with material consequences. Reviewers should be given clear explanations of transaction semantics and supporting evidence. Adding a confirmation box for users to click, without explaining the actual asset changes, does not create a reliable security boundary. Summary and Recommendations The most concerning aspect of the “AI crisis” is the efficiency gap between vulnerability discovery, attack implementation, and security hardening. Smart contracts offer a short path from vulnerability to financial loss, so this gap can quickly translate into asset risk. Cryptographic research and migration operate on longer cycles. When a critical security assumption changes, the consequences can span multiple systems and the foundations of business operations. The reality of AI-driven vulnerability discovery demands that defenders take AI’s end-to-end vulnerability discovery and exploitation capabilities, as well as its cryptanalysis capabilities, seriously. Using AI to counter AI may be the only way to address this crisis. We should move quickly to establish AI-powered continuous auditing, security monitoring, and incident response. References The links below point to the original papers, specifications, and official materials consulted or checked for this report. Dynamic webpages and repositories were accessed on October 26, 2026; paper versions are specified in the individual entries. Experiments described by the researchers were not independently rerun for this report. [1] Anthropic / MATS / Anthropic Fellows. AI agents find smart contract exploits, December 2025. https://www.anthropic.com/research/smart-contracts Supports the figures for SCONE’s 405 tasks, Best@8, simulated profits, and the screening experiment involving 2,849 new contracts. [2] Anthropic. Measuring LLMs’ ability to develop exploits, May 22, 2026. https://www.anthropic.com/research/exploit-evals Supports the 12 updated tasks, Best@8, and approximately $35 million in simulated profits. [3] Wang et al. EVMbench: Evaluating AI Agents on Smart Contract Security, arXiv:2603.04915v1, March 5, 2026. https://arxiv.org/html/2603.04915v1 All EVMbench figures in this report use this version. [4] Anthropic. scone-bench, repository documentation, accessed October 26, 2026. https://github.com/anthropics/scone-bench Explains the difference between the 417-task dataset and the earlier 405-task snapshot. Content on the main branch may continue to change. [5] Anthropic. Discovering cryptographic weaknesses with Claude, July 28, 2026; the page records an edit on July 29. https://www.anthropic.com/research/discovering-cryptographic-weaknesses Provides the researchers’ account of HAWK, reduced-round AES, and the discovery process. [6] Straznickas & Weis. HAWK-n Key Recovery Reduces to SVP in Dimension n/2 + 1, 2026. https://www.anthropic.com/document/hawk_key_recovery.pdf Supports the reduction structure, gate-count costs, and HAWK-256 experiments. See the abstract and Section 6. [7] HAWK Design Team. Hawk official website and withdrawal announcement dated July 29, 2026. https://hawk-sign.info/ Confirms the scheme authors’ formal response to the new attack and their withdrawal, independently of the model vendor’s account. [8] Nasr & Carlini. Cryptanalysis of 7-Round AES via the Algebraic Structure of its S-box, 2026. https://www.anthropic.com/document/aes_mobius_bridge.pdf Supports the seven-round scope, data complexity of 2^105, time complexity of 2^89.3–2^91.4, and the statement that the attack was not executed end to end. [9] Fluri et al. CryptanalysisBench: Can LLMs do Cryptanalysis?, arXiv:2607.18538v2, July 29, 2026. https://arxiv.org/html/2607.18538v2 Supports the 191 tasks, validation through security games, and classification of design, specification, and implementation flaws. See especially Sections 3, 4, and 5.4. [10] Gohr. Improving Attacks on Round-Reduced Speck32/64 using Deep Learning, CRYPTO 2019. https://eprint.iacr.org/2019/037 Documents earlier advances in neural distinguishers and attacks on an 11-round reduced variant. [11] Wenger et al. SALSA: Attacking Lattice Cryptography with Transformers, arXiv:2207.04785v2, revised in 2023. https://arxiv.org/html/2207.04785v2 Supports machine-learning analysis of LWE instances with sparse binary secrets and the associated parameter limitations. [12] Prouff et al. Study of Deep Learning Techniques for Side-Channel Analysis and Introduction to ASCAD Database, 2018, with subsequent revisions. https://eprint.iacr.org/2018/053 Supports research on deep-learning side-channel attacks against masked AES and reproducible experiments. [13] Hitaj et al. PassGAN: A Deep Learning Approach for Password Guessing, arXiv:1709.00440v3, 2019. https://arxiv.org/abs/1709.00440v3 Supports candidate-generation methods that learn human password distributions. [14] Ethereum. EIP-712: Typed structured data hashing and signing. https://eips.ethereum.org/EIPS/eip-712 The specification explicitly states that it does not itself include replay protection. [15] Pornin. RFC 6979: Deterministic Usage of the Digital Signature Algorithm (DSA) and Elliptic Curve Digital Signature Algorithm (ECDSA), 2013. https://www.rfc-editor.org/rfc/rfc6979.html Supports the standard method for deterministic signing nonce generation. [16] NIST. SP 800-57 Part 1 Revision 5: Recommendation for Key Management, 2020. https://csrc.nist.gov/pubs/sp/800/57/pt1/r5/final Provides the baseline framework for key management and security-strength comparisons. [17] NIST. FIPS 203: Module-Lattice-Based Key-Encapsulation Mechanism Standard, August 13, 2024. https://csrc.nist.gov/pubs/fips/203/final The ML-KEM standard. Official errata should also be checked during deployment. [18] NIST. FIPS 204: Module-Lattice-Based Digital Signature Standard, August 13, 2024. https://csrc.nist.gov/pubs/fips/204/final The ML-DSA standard. [19] NIST. FIPS 205: Stateless Hash-Based Digital Signature Standard, August 13, 2024. https://csrc.nist.gov/pubs/fips/205/final The SLH-DSA standard. [20] OpenAI Docs. Safety in building agents, accessed October 26, 2026. https://developers.openai.com/api/docs/guides/agent-builder-safety Provides general guidance on prompt injection, untrusted input, and tool-permission risks. This report does not rely on the page’s product-specific configuration recommendations. [21] Biryukov et al. RFC 9106: Argon2 Memory-Hard Function for Password Hashing and Proof-of-Work Applications, 2021. https://www.rfc-editor.org/rfc/rfc9106.html Supports memory-hard password derivation and the need to assess implementation parameters for the intended use case. [22] NIST. FIPS 197: Advanced Encryption Standard (AES), updated in 2023. https://csrc.nist.gov/pubs/fips/197/final Supports AES parameters and standard round counts. [23] Shor. Polynomial-Time Algorithms for Prime Factorization and Discrete Logarithms on a Quantum Computer. https://arxiv.org/abs/quant-ph/9508027 Provides theoretical background on quantum algorithms for factoring and discrete logarithms. [24] Grover. A fast quantum mechanical algorithm for database search. https://arxiv.org/abs/quant-ph/9605043 Provides theoretical background on quantum unstructured search.(Consensus expectations: Consensus for 2026Q3 expects EPS of -0.99 on revenue of USD 8.128 million.)

25/100·CNeutral+26

news10/28news

THE BLOCK: Crypto’s legislative outlook could shift sharply as a result of next week's midterm elections, with Democratic control of the House or Senate, or both, at play. A Democratic-controlled Congress could mean more oversight, a tougher path for the Clarity Act, and renewed scrutiny of Trump-family-led World Liberty Financial.

72/100·B+Short+3