Google Project Zero cuts its bug deadline to 30 days by 2028, saying AI lets attackers exploit patches at once
Project Zero dropped its 30-day patch grace period and will shrink disclosure deadlines from 90 to 60 to 30 days, citing LLM-speed attackers. Patch windows will tighten.
On November 13, 2026, Google Project Zero replaced its "90 + 30" disclosure rule with a flat "Full X" deadline. Bug details now go public X days after the vendor is told, with no extra 30 days for patches to roll out. X is 90 today, drops to 60 on February 2, 2027, and to 30 on January 3, 2028. The team says large language models (LLMs) now let attackers find bugs and reverse-engineer patches so fast that the old buffer no longer protects anyone. Anyone who installs updates should expect less time between a patch shipping and attackers using it.
What happened, in plain English
Project Zero is Google's team of vulnerability hunters. They look for serious security bugs in widely used software (operating systems, browsers, phones, chips), report each one privately to the vendor, and publish the details after a deadline. That deadline pushes vendors to fix bugs instead of sitting on them.
Since 2021 the rule has been "90 + 30": the vendor gets 90 days to fix the bug. If it ships a fix in time, Project Zero waits up to another 30 days, the "soak time", so customers can install the update before the technical details come out.
The new policy, which Project Zero calls "Full X", removes that second window. Details are published a fixed number of days after the report, whether or not a patch exists and whether or not people have installed it.
The reason is AI. According to Project Zero, "LLMs have increased the speed at which attackers can find and exploit vulnerabilities," and they have "decreased the time it takes for attackers to reverse engineer and exploit a vulnerability." The team also says it is now "finding several vulnerabilities a week in targets where we once found a few vulnerabilities a year."
An analogy: the old rule was like a locksmith quietly telling a landlord about a weak lock, then waiting an extra month after the new locks arrived before writing about it. The extra month only helps if burglars can't tell from the new lock what was wrong with the old one. Project Zero now believes they can, almost at once.
Are you affected? What to do now
Most readers don't have to do anything about Project Zero itself. The policy mostly affects the vendors whose products it tests. What it tells you is how quickly a released patch can become an attack, and that applies to everyone who installs updates.
Who should care most
- Sysadmins and endpoint teams who patch Windows, macOS, Linux, browsers, mobile devices and firmware.
- Security and risk leads whose patch policies allow "30 days for high, 90 days for medium" or similar.
- Product and engineering teams that ship software and may receive Project Zero reports.
Checklist
- Measure your real patch time. For browsers, operating systems and internet-facing devices, check how many days pass between a vendor's release and full deployment. If the answer is "weeks", that gap is the risk Project Zero is describing.
- Speed up the fast lane. Let browsers and other auto-updating software update themselves. Keep a tested emergency process for patching within days, not at the next monthly window.
- Don't wait for a severity score. Project Zero says severity tells you how much harm a bug can do, not how likely attackers are to find it. Treat "low/medium" fixes on exposed systems as more urgent than before.
- Prepare non-patch mitigations. For devices you can't update quickly (appliances, OT, legacy systems), list in advance what you can do instead: disable the affected feature, restrict network access, add monitoring.
- Ask your suppliers. Ask key vendors how they plan to ship fixes on shorter timelines, and whether they publish interim mitigations when a fix isn't ready.
- If you build software: assume a bug report from a capable researcher may be public within 90 days now, and within 30 days from January 2028. Budget engineering time for that, and plan how you would protect customers before a patch is ready.
Actively exploited bugs are not covered by Full X. Project Zero says they follow Google's separate application-security disclosure policy.
The expert view
What actually changed. There are two changes. First, the deadline now counts only from the report, so there is no automatic post-patch embargo. Second, Project Zero dropped the special cases it used to negotiate: the grace period and the variant policy are gone. The remaining exceptions are in-the-wild exploitation (handled under Google's broader policy), bugs already made public elsewhere, and a vendor asking for earlier disclosure. Project Zero will keep tracking whether fixes land and will say when they haven't, but fix status no longer affects timing. It says it has occasionally granted exceptions, Spectre and Meltdown among them, and will keep doing so.
Why soak time stopped working. Soak time assumed that a patch hides its bug: attackers had to compare the patched and unpatched code and work out the flaw themselves, and that took long enough for most users to update. Project Zero's view is that LLM-assisted reverse engineering has largely removed that delay, and that "zero patch-to-exploitation time is a reasonable, though imperfect, model of attacker behavior." If that's true, holding back details after the patch protects users very little and mostly slows defenders. Our November 12 story on Fortinet's AI-found flaws made the same practical point: patch before attackers compare the old and new code.
Discovery is faster too. The second argument is about discovery. If AI-assisted tools let more people find the same bugs, the chance that someone malicious finds a reported bug independently rises. Project Zero says this makes a newly reported bug a more immediate risk than one reported a year ago, so long embargoes make less sense. This matches what this site has covered in recent weeks: vendors and researchers crediting AI tooling with large batches of findings, and threat-intel teams describing attackers running AI agents against exposed systems.
What's new and what isn't. Fixed disclosure deadlines are not new. Project Zero has used them for years, and plenty of coordinated-disclosure programs work the same way. What's new is a major research team rewriting its policy explicitly because of AI, and putting dates on it. The gradual timeline reflects vendor feedback: Project Zero says an immediate switch risked hurting users and could have led vendors to deprioritize lower-severity bugs or stop trying to meet deadlines.
What's still unknown.
- The post doesn't say how Full X applies to bugs reported before November 13.
- Project Zero gives no measured data on patch-to-exploit times. It describes "zero" as a working model.
- It's not yet clear whether other disclosure programs, bug-bounty platforms or national coordinators (CERTs) will follow. If they do, the shorter deadlines will spread across the industry. If they don't, Project Zero reports will stand out as unusually fast.
- Hardware and embedded vendors face the hardest test. Project Zero tells vendors to prepare for imminent disclosure, including mitigations that don't depend on a patch.
Bottom line for defenders: treat the date a patch is released as the date attackers start reading it.
Official sources
- Google Project Zero: Disclosure Policy Change: 2026 Edition (November 13, 2026)
- Coverage: Cloud Security Alliance, CISO Daily Briefing, November 15, 2026
- Coverage (background on Project Zero's 2025 Reporting Transparency trial): TechRadar, July 31, 2025
- Related on AI Cyber Brief: Fortinet details critical firewall and FortiManager flaws its AI testing found