Anthropic AI Scanner Targets Crypto Wallet Security Bugs
Anthropic opened its new OSS Scanner on 8 October 2026, and crypto projects including Nethermind, ZEUS and VirtEngine applied the next day. The tool uses advanced AI models to hunt bugs before attackers find them, a step that matters to anyone who trusts a crypto wallet with real money.
What actually happened
Anthropic launched OSS Scanner on 8 October 2026, offering free vulnerability reports from its strongest AI models, including Claude Mythos, according to Cointelegraph. Crypto firms applied the next day, 9 October 2026. Ethereum client builder Nethermind requested a full repository scan. Bitcoin and Lightning wallet ZEUS asked for checks on payment flows, private keys and Lightning Service connections. VirtEngine, a Cosmos based decentralized cloud marketplace, also applied. Anthropic said the reports exist 'to give open-source projects the largest defensive advantage.' No pull requests had been merged as of publication.
How we got here
OSS Scanner follows Anthropic's earlier Project Glasswing, which also scanned open-source code but needed slow human review before sharing findings. Crypto has had a rough year on security. Bitcoin swap service Boltz halted operations in August 2026, saying attackers patched exploits faster than its team could respond. Payment platform PayPerQ reported repeated attacks it believed were AI assisted. Anthropic itself warned on 8 October 2026 that AI currently helps attackers more than defenders, since exploitation gets cheaper while fixing bugs still needs human effort.
Why this matters for you
For everyday wallet users, this could mean bugs get caught and patched before hackers drain funds, a fix you never see but benefit from. For developers building wallets or Web3 identity tools like bonuz, free access to frontier model audits lowers the cost of serious security review. But access is not open to everyone yet. Anthropic picks applicants case by case, weighing how critical a project is, how exposed it is to remote attacks, and how many users rely on it. That selection could decide which wallets stay resilient and which ones wait their turn.
The bigger question
If frontier AI can spot vulnerabilities faster than any human team, who should decide which projects get protected first? Anthropic currently reviews applicants case by case instead of opening the scanner to everyone at once. That raises a wider question: can defensive AI access stay fair, or will it become another resource only well connected projects can secure quickly?
What to watch
Watch the OSS Scanner GitHub repository for merged pull requests from Nethermind, ZEUS and VirtEngine, none confirmed as of 9 October 2026. More applications are expected from AI agent and wallet security teams as the program expands. Bonuz will keep tracking how tools like this affect wallet safety and the hardware tied to Web3 identity.
