Business IEA Says Data Center Power Demand Could Double to 1,000 TWh by 2030, With AI Supplying Three Quarters of the Growth Chatbots ChatGPT Now Transcribes as You Talk: OpenAI Removes the 10-Minute Cap and Streams Words Live on Every Plan Business Ex-Alibaba Executive's Everful Raises $48.5 Million to Put AI Agents in Charge of Sourcing From 30,000 Factories AI Agents Google Built a Mirror Dimension of the Internet to Train Its Chrome AI Agents Without Breaking the Real Web Productivity Canva Finally Adds an Off Switch for Its Generative AI Features, and a Filter to Hide AI-Made Templates Business Anthropic Begins Trading Today: 980 Million Shares at $83, a $1.5 Trillion Valuation and the Largest IPO on Record News Google Books Blue Origin New Glenn to Launch TPU-Carrying AI Satellites in Early 2029 Security Hackers Used the AI Penetration Tool ARTEX to Disrupt European Commission Systems for Over Ten Days Business IEA Says Data Center Power Demand Could Double to 1,000 TWh by 2030, With AI Supplying Three Quarters of the Growth Chatbots ChatGPT Now Transcribes as You Talk: OpenAI Removes the 10-Minute Cap and Streams Words Live on Every Plan Business Ex-Alibaba Executive's Everful Raises $48.5 Million to Put AI Agents in Charge of Sourcing From 30,000 Factories AI Agents Google Built a Mirror Dimension of the Internet to Train Its Chrome AI Agents Without Breaking the Real Web Productivity Canva Finally Adds an Off Switch for Its Generative AI Features, and a Filter to Hide AI-Made Templates Business Anthropic Begins Trading Today: 980 Million Shares at $83, a $1.5 Trillion Valuation and the Largest IPO on Record News Google Books Blue Origin New Glenn to Launch TPU-Carrying AI Satellites in Early 2029 Security Hackers Used the AI Penetration Tool ARTEX to Disrupt European Commission Systems for Over Ten Days

Anthropic Cuts Its Bug Disclosure Window to 30 Days as AI Finds Vulnerabilities Faster Than Vendors Can Patch

Anthropic Cuts Its Bug Disclosure Window to 30 Days as AI Finds Vulnerabilities Faster Than Vendors Can Patch

Anthropic will publish unpatched high and critical vulnerabilities after 30 days instead of 90 starting January 1, 2027, saying AI-driven bug discovery has broken the old disclosure math. Escalation and patch-detail windows tighten to 15 and 30 days while the 7-day clock for actively exploited bugs stays unchanged.

Anthropic said Monday it is cutting the time it gives software vendors to fix serious security flaws from 90 days to 30, arguing that AI can now find and independently rediscover the same bugs far faster than the old disclosure clock assumed. The new timelines apply to vulnerabilities reported on or after January 1, 2027.

Under the updated coordinated vulnerability disclosure policy, high- and critical-severity issues will be published 30 days after being reported if no fix has shipped, down from 90. All other issues move to a 45-day window. If a maintainer does not respond at all, Anthropic will escalate to an external vulnerability coordinator after 15 days instead of 30, and once a patch is available, full technical details will follow after 30 days rather than 45. The seven-day disclosure window for bugs already under active exploitation is unchanged, and reports already in progress keep their existing deadlines.

The company said the 90-day standard, which traces back to Google's Project Zero, was written for a world where fixes moved at human speed and the same flaw was unlikely to be found twice. Neither assumption holds anymore, Anthropic argued: models can help draft fixes, and a bug one model finds is increasingly likely to be found by another as rediscovery rates climb.

The company's own tooling illustrates the shift it is responding to. Claude Mythos Preview can reverse-engineer exploits on closed-source software, and the free OSS Scanner has surfaced more than 29,000 candidate vulnerabilities in open-source projects, figures Anthropic reported itself and which have not been independently audited. On the same day, security firm Forescout reported that AI models found 66 previously unknown flaws in critical infrastructure code in about 90 hours, work its researchers estimated would have taken humans roughly nine months.

The new policy also clears a backlog: high- and critical-severity bugs Anthropic reported more than 90 days ago that are still undisclosed will get one further 30-day window, after which the reports will be published on the company's dashboard starting January 1, 2027.

Anthropic carved out exceptions for software that is hard to update, such as critical infrastructure, and for flaws requiring specification or RFC changes, where it will agree a remediation timeline with the vendor. Raw AI-generated findings that have not received human review are exempt from the policy entirely.

The move puts Anthropic ahead of the industry benchmark it is departing from: Google's Project Zero still holds its 90-day deadline. If AI-hunted bugs keep arriving at the pace Forescout described, the question for the rest of the industry is not whether the 30-day clock spreads, but how soon vendors get the resources to keep up with it.

Comments (0)

Log in to join the discussion

Log In

No comments yet