Friday, November 20, 2026

Top 5 This Week

Related Posts

AI finds flaws too quickly: Anthropic and Google end 90-day grace period before disclosure


Anthropic will only give publishers 30 days to correct serious flaws that it reports to them, compared to 90 so far. Three days earlier, Google Project Zero had announced a comparable, but more gradual decline.

For years, the computer security community has followed an unspoken rule: a researcher who finds a flaw discreetly notifies the publisher, then gives them 90 days before making it public. This three-month delay should allow the patch to reach everyone before hackers know of the flaw and use it.

However, in a post published this Monday, November 16, Anthropic, the company behind the Claude artificial intelligence, decided to reduce this deadline to 30 days for high or critical severity flaws, and to 45 days for others. The new rules apply to reports made from January 1, 2027.

And the underlying reason for this upheaval in the principles of coordinated and responsible disclosure? AI.

A flaw no longer remains secret for three months, estimates Anthropic

Anthropic justifies this change by the rapid progress of AI models, which have become much better at finding flaws. According to the company, a breach spotted by an AI has an increasing chance of being found by another, well before the 90-day deadline, which greatly increases the degree of IT security risk.

Anthropic also says it is already seeing more rediscoveries. Google starts from the same observation.

1,000 AI and 30 hours to verify Fermat’s theorem: Anthropic’s coup to launch Claude Fable 5.2 // Source: Numerama / Anthropic

On November 13, Project Zero, the Google bug-hunting team that popularized the 90-day rule, announced it would move to 60 days on February 2, 2027, then to 30 days on January 3, 2028, whether the patch is ready or not and regardless of the severity of the flaw. The team is also removing the additional 30 days it has allowed since 2021 to deploy a patch.

This pace particularly weighs on maintainers, who maintain free software, and Anthropic recognizes this. She points out that she already attaches a correction to her reports when possible. It also offers maintainers of critical open source projects OSS Scanner, a free service that regularly screens their code for its models.

In short, the company whose models accelerate the hunt for flaws also offers to put them at the service of those who must correct them. Software that is difficult to update, such as that for critical infrastructure, will have a schedule negotiated on a case-by-case basis.

Flaws already exploited by hackers retain their regime: they remain subject to the same seven-day deadline as before. The new deadlines also do not apply to OSS Scanner finds that no human has yet reviewed.

All the tech news at a glance

All the tech news at a glance

Add Numerama to your home screen and stay connected to the future!


Shredder Smith
Shredder Smith
Shredder Smith is the lead curator and digital persona behind topaitools4you.com, an AI directory dedicated to "shredding" through industry hype to identify high-utility software for everyday users. Smith positions himself as a blunt, no-nonsense reviewer who vets thousands of emerging applications to filter out overpriced "wrappers" in favor of tools that offer genuine ROI and practical productivity. The site serves as a watchdog for the AI gold rush, providing categorized rankings and transparent reviews designed to help small businesses and creators navigate the crowded tech landscape without wasting money on low-value tools.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Popular Articles