[ctx]ctxseccontext security: attacks on the harness and context layer of AI agents, plus exploited-now news, hunting and playbooks.

← AI · markdown

confirmed (Bitcoin Core security policy page); merge date Oct 22 per Bitcoin.com News and TSN; announced by developer Antoine Poinsot on the bitcoin-dev mailing list per Cointelegraph

Bitcoin Core drops embargoes for bugs 'discoverable with a simple LLM prompt': fixes ship to every maintained version at once, with an advisory

Bitcoin Core's disclosure policy now has an 'easily discoverable' class: vulnerabilities 'that can be discovered with minimal effort, such as through a simple LLM prompt'. For these, the project will publish a point release with the fix for each maintained major version as soon as possible, together with a security advisory, skipping the usual severity-based embargo. Under the normal rules, Medium and High bugs stay undisclosed until about two weeks after the last affected release reaches end of life, which is roughly a year.

Poinsot's reasoning, as reported by Cointelegraph: once an LLM can easily rediscover a bug, 'the embargo period achieves little but it delays the availability of a fixed release.' The trade-off developers acknowledged is less time to test fixes before release. Other projects that run long embargoes face the same question as AI-assisted bug hunting spreads.

Sources: Bitcoin Core: Security Advisories and policy · Cointelegraph (Oct 23) · CoinDesk (Oct 23)