In a major change, AI company Anthropic and Google's Project Zero are cutting down the time they give software makers to fix security flaws. Both companies are moving away from the old 90-day rule to keep users safer, as artificial intelligence (AI) makes it much faster to find bugs.Anthropic's new 30-day ruleOn November 16, 2026, Anthropic shared an update about how it handles security bugs.
The company is reducing the time it gives software makers to patch severe bugs from 90 days down to just 30 days. For less severe issues, companies will have 45 days.
These new rules will kick in for bugs reported to software makers on or after January 1, 2027.Anthropic is also speeding up other parts of the process. If a software maker ignores a bug report, Anthropic will now reach out to an external coordinator after 15 days instead of 30.
Also, once a software update is available, Anthropic will share the full technical details of the bug after 30 days, rather than waiting for 45 days.If hackers are already using a bug to attack people, companies still only get a strict 7-day window to fix it. For older bugs that were reported over 90 days ago and are still unfixed, Anthropic is giving developers a final 30-day warning before putting the details on its public dashboard.
Google Project Zero cuts deadlinesGoogle’s security team, Project Zero, is making similar moves. Currently, they give companies 90 days to fix bugs, plus an extra 30-day grace period.
But starting February 2, 2027, Project Zero will lower the time limit to 60 days. By January 3, 2028, it will drop to just 30 days.Google is also completely removing the extra 30-day grace period.
The reason is that attackers today can figure out how to break patches almost immediately after they are released.Why the big rush? The main reason for these shorter deadlines is AI.
Anthropic noted that AI models can now spot and even help fix vulnerabilities much faster than humans ever could. In the past, companies were given 90 days because it was rare for two different people to find the exact same bug independently.Today, if one AI tool finds a security flaw, another AI tool will likely find it very soon after.
Because of this, holding back bug details for months is no longer safe. Some bugs, like those in very hard-to-update software, cannot be fixed in 30 days.
Anthropic said it will work with vendors on special timelines for those rare cases.To help developers deal with the faster pace, Anthropic is offering a few free tools. Maintainers of important open-source projects can use a free service called OSS Scanner to automatically check their code for bugs.
They can also apply for six months of free access to Claude Ultra, Anthropic's advanced AI tool, to help them fix issues quickly. Finally, security workers can apply for a Cyber Verification Program, which adjusts Claude's safety filters for legitimate defensive work.Ultimately, these changes mean that software makers will have to roll out security updates much faster.
While it puts more pressure on developers, it means better and faster security for everyday consumers and their devices.- Ends

