Skip to content
by InfinyAI ES

IA en un minutoNewsSecurity

Security

Anthropic shortens from 90 to 30 days the time it takes to disclose serious security flaws it discovers

IA en un minuto newsroom · Editor: Jon Elgezabal

In 30 seconds

Anthropic is lowering to 30 days, down from 90, the margin it gives manufacturers before publishing a serious security flaw it has found, and 45 days in all other cases, starting in January 2027. This is because AI models find bugs better and better and another is likely to hit the same one. It keeps seven days for flaws already being exploited and will agree a timeline case by case for hard-to-deploy software.

Anthropic is going to give manufacturers less time to fix the security flaws it reports to them before making them public. It announces this on its website.

The company is shortening the window in its coordinated vulnerability disclosure policy (the rules for warning about a flaw before making it public): it goes from 90 to 30 days for high- and critical-severity flaws, and to 45 days for the rest. It will apply to flaws it reports to vendors and maintainers on or after January 1, 2027; reports already in progress on that date keep their current deadline.

There are two other changes. If a maintainer does not answer the first report, Anthropic will try to bring in an external coordinator after 15 days rather than 30. And once a patch is published it will wait 30 days, instead of 45, before sharing the full technical details. In addition, the serious flaws it reported more than 90 days ago that are still not public already exceed the deadline in the current policy: Anthropic is telling the maintainers it is giving them 30 more days, after which it will publish those reports on its dashboard under the new policy, starting January 1, 2027.

The reason, according to Anthropic, is that AI is improving so fast that the earlier rules, designed for human-paced fixes and a low risk of someone else rediscovering the same flaw, no longer hold: one model can find a flaw and another can find it too, and the company says it is seeing more rediscoveries. For software that is hard to deploy, such as critical infrastructure, and for flaws that require changing a specification, it will agree on its own timeline with the manufacturer to fix it.

The seven days for flaws already under active exploitation stay the same, and the policy does not apply to AI findings that have not had human review, such as those from OSS Scanner. Anthropic acknowledges that shorter windows may put pressure on maintainers and says it will keep including guidance in each report and, where it can, a patch or a mitigation.

Why it matters · analysis and opinion

The change shifts the pressure onto whoever maintains the software. With a month to fix a serious flaw, a project without a dedicated team, especially an open-source one, will need someone watching the security inbox and a process for patching fast. Anthropic acknowledges that burden and offers guidance, patches when it can and free access to OSS Scanner for maintainers of critical projects. It also points to a deeper shift: if several models can find the same bug, keeping it secret for months stops being realistic and the advantage goes to whoever fixes first. For any company that uses that software, the consequence is to install updates as soon as they come out, because the window between the patch and the public details is shrinking.

Official source: Anthropic · Written with the help of AI: how we make the news

What does it mean for your business? · analysis and opinion

For most small businesses and freelancers in Spain and Latin America, this does not change anything directly, because Anthropic is not going to contact them. Who it does affect is small companies that build or maintain software, especially open-source projects: if they receive a notice from Anthropic on or after January 1, 2027, they will have 30 days to fix a high- or critical-severity flaw and 45 days for the rest. If they do not respond to the first notice, Anthropic will look for an external coordinator after 15 days, and the full technical details will be published 30 days after a patch is available.

Our advice for these companies is to have a visible security email or channel that is reviewed daily, and to decide now who responds to these notices and within how much time. If the software is hard to deploy, it is a good idea to ask early for an agreed timeline, because Anthropic says it will negotiate one in these cases. Maintainers of critical open-source projects can apply for OSS Scanner, a no-cost service that scans their code with Anthropic models.

For other businesses, the sensible move is to install security updates as soon as they come out. If the flaws become public earlier, the time between when a patch exists and when anyone can know the details gets shorter, and the risk of leaving systems unpatched grows.

AI in your inbox, every day or every Friday

The stories that matter, each one in a minute. With the source for every one.

Choose one or both:

Sign up and you are in: the daily arrives every night and the weekly on Friday mornings. You can unsubscribe from any email.