Incident sheet · /security
- System
- Security & Risk
- Filed
- Read
- 4 min · 749 words
- Severity
- Routine
Google and Anthropic Compress Vulnerability Disclosure Windows
Google and Anthropic are cutting vulnerability disclosure windows to as little as 30 days, forcing enterprise teams to prepare for fixes that arrive before patches exist.

Event log
- Google will cut the remediation window from 90 to 60 days in February 2027 and to 30 days by January 2028.
- Anthropic starts 30-day disclosures for high and critical model-found flaws and 45 days for others on January 1, 2027.
- Anthropic will reduce post-patch technical publication from 45 to 30 days and retain seven days for actively exploited flaws.
- Google’s current 90+30-day policy includes 30 days of soak time; the overhaul removes grace periods and fix-verification negotiations.
- A Google Threat Intelligence Group report says LLM frameworks can automate vulnerability research and exploitation with no human in the loop.
Google will cut vendor remediation deadlines from 90 days to 30 by January 2028, while Anthropic will start 30-day disclosures for high-severity and critical flaws on January 1, 2027.
What changes in the disclosure deadlines?
Google Project Zero and Anthropic announced the changes independently this week. For flaws found by its researchers, Google currently uses a 90+30-day policy: 90 days for vendors to remediate, followed by 30 days of patch soak time.
Beginning in February 2027, Google will reduce the remediation window to 60 days, then tighten it to 30 days by January 2028. The final target removes 60 days, or two-thirds, from Google’s current remediation period.
The Google overhaul also removes grace periods and negotiations over fix verification. Project Zero tied those changes to surging vulnerability volume, which has made manual tracking unmanageable.
Starting January 1, 2027, Anthropic’s policy for flaws discovered by its frontier models will allow 30 days—roughly one month—for high- and critical-severity issues and 45 days, or about six weeks, for all other vulnerabilities. It will also reduce the post-patch technical-publication delay from 45 days to 30 days.
Complex environments and critical infrastructure can still receive tailored schedules, while actively exploited flaws retain a seven-day window.
Takeaway: Vendors will have less time to negotiate, test and publish repairs.
Why are the labs moving faster?
AI drives both changes. Google and Anthropic say frontier models can autonomously discover, reverse-engineer and weaponize software flaws within hours. Their conclusion is practical: a long embargo gives attackers a reusable blueprint while giving defenders false assurance that the flaw remains hidden.
Natalie Silvanovich, a security engineer at Google Project Zero, said: “A vulnerability reported today presents a much higher immediate risk to users than one reported last year.” She said large language models have changed the risk profile and made vendor negotiations over fixes and grace periods increasingly unmanageable.
Anthropic’s Frontier Red Team reported a rising rediscovery rate: a flaw one model can find is increasingly likely to be found by another as capabilities improve. That independent duplication erodes any advantage from a prolonged embargo.
Michael Moore, Anthropic’s cybersecurity lead, said in a LinkedIn post: “Frontier models have become much better at finding vulnerabilities.” He expects shorter disclosures to help defenders fix flaws faster, but also expects other models to reproduce each other’s findings.
Takeaway: AI has sharply reduced the useful life of a secret vulnerability.
Can vendors meet the new deadlines?
The new schedule transfers pressure to software vendors and the enterprises that buy their products. Ramy Rahman, senior principal solutions engineer at ArmorCode, said the changes place greater pressure on vendors to address vulnerabilities.
His practical advice is to assess whether compensating controls can reduce exposure when a disclosure arrives before a patch. Teams need a repeatable process for evaluating exposure, applying available mitigations and tracking remediation rather than waiting for a full fix.
Neil Osnato, founder of Persistence Analytics Group, separates disclosure from remediation: one transmits information, while the other changes a running system. Faster disclosure, he said, does not necessarily mean faster risk resolution.
Systems may remain exposed because of:
- compatibility testing;
- third-party dependencies;
- maintenance windows;
- legacy infrastructure; and
- the consequences of taking a critical system offline.
Osnato described operational deployment as a separate institutional decision, not an automatic consequence of a shorter reporting clock.
Takeaway: A faster clock for reporting does not guarantee a faster fix in production.
What does automated exploitation change?
Google warned this week that AI threats had crossed a significant threshold. A new Google Threat Intelligence Group report says threat actors use large-language-model frameworks to automate vulnerability research and exploitation with no human in the loop.
That development raises the risk from n-day vulnerabilities, meaning known flaws that often remain unpatched for months or years in enterprise environments. Automated discovery and exploitation can compress the interval between public disclosure and active abuse.
Jason Elrod, CISO at MultiCare Health System, said traditional patching cycles are becoming obsolete. He warned that some fixes must happen within hours rather than days or weeks.
AI labs frame shorter disclosure windows as a way to improve industry-wide resilience against machine-speed attacks. The mechanism is straightforward: publish sooner, shorten attackers’ opportunity and give defenders time to act. The constraint is equally clear: disclosure cannot remove compatibility, dependency or deployment delays.
Takeaway: Enterprise defenders now face an automated race that faster reporting alone cannot close.
via techtarget.com (Source)
More from Amara Osei
Show full bio
Correspondent covering regulation, policy and their impact on enterprise information technology and the leaders who run at CIO Monitor.
188 articles
The newsroom digest
One email a day with the stories that matter. No noise.
Cross-references
- Google Project Zero Cuts Bug Disclosure Windows to 30 Days by 2028
- FedRAMP Sets Dec. 7 Deadline for AI-Driven Vulnerability Rules
- Google Finds AI Agents Turning Patches into Working Exploits Autonomously
- Google extends Gemini temporary chat retention to two months
- Amazon, Meta, Google and Microsoft retreat from data center secrecy