LIVE FEED
CRITICAL SGLang RCE Exploit Breaches NEAR AI, Hits 18 Agents // FIRST LOOK F-Secure Releases AI Harms Taxonomy Mapping 26 Consumer Risks // CRITICAL Cursor Marketplace Trojanized via Stolen Fireworks AI Key // CRITICAL Cairn AI Framework Hits Semiconductor and Defense Orgs // HIGH TypoPrompting: 186 Fake AI Domains Spread ClickFix Malware // CRITICAL Post-Training Supply Chain Attacks Backdoor AI Coding Agents // HIGH Google Project Zero Cuts Disclosure Window as LLMs Speed Up Exploits // HIGH Unsanctioned AI Agents Expose Five Enterprise Identity Security Gaps // HIGH OpenAI Agents Gain Root Access, Conceal Activity in CI Clusters // FIRST LOOK AWS Launches A2A Protocol for Interoperable Multi-Agent Systems //
ATLAS OWASP HIGH Significant risk · Prioritise patching LLM Security Research Regulatory Industry News RELEVANCE ▲ 7.2

Google Project Zero Cuts Disclosure Window as LLMs Speed Up Exploits

SOURCE Google Project Zero ↗
TL;DR HIGH
  • What happened: Project Zero is slashing disclosure timelines to 30 days by 2028 because LLMs now dramatically accelerate exploitation.
  • Who's at risk: Software vendors and their end users are most exposed, as compressed timelines leave less time to patch before vulnerabilities are publicly disclosed.
  • Act now: Accelerate internal patch development and release pipelines to meet sub-60-day remediation targets · Implement in-product mitigations or virtual patching to protect users while full patches are prepared · Monitor patch releases for reverse-engineering exposure and assume near-zero delay before attacker exploitation
Google Project Zero Cuts Disclosure Window as LLMs Speed Up Exploits

Overview

Google Project Zero has announced a fundamental revision to its vulnerability disclosure policy, explicitly citing the impact of large language models (LLMs) on the attacker threat landscape. The new ‘Full X’ policy removes grace periods and most special-case provisions, transitioning from the current 90-day window down to 60 days in February 2027 and 30 days by January 2028. This is one of the first major institutional disclosure policy changes directly attributed to AI-accelerated offensive capabilities.

Technical Analysis

Project Zero identifies three distinct ways LLMs have shifted the calculus of coordinated disclosure:

  1. Accelerated vulnerability discovery: LLMs are enabling researchers — and attackers — to find vulnerabilities at dramatically higher rates. Project Zero notes it is now finding several vulnerabilities per week in targets where it previously found a few per year.

  2. Near-zero patch-to-exploitation time: LLMs can reverse engineer bug details directly from patch diffs, meaning the traditional ‘patch soak’ period — where a fix is released but not yet widely deployed — no longer provides meaningful user protection. The assumption that patch contents remain opaque to attackers is no longer valid.

  3. Increased simultaneous discovery risk: The probability that multiple independent parties, including malicious actors, discover the same vulnerability concurrently has increased, raising the stakes of any delay in disclosure.

The policy removes negotiated exceptions such as grace periods and variant policies, replacing them with a simple timer-based model designed to be automatable by both Project Zero and vendors.

Framework Mapping

This development maps most directly to AML.T0047 (AI-Enabled Product or Service) — LLMs are being used as offensive tools to accelerate vulnerability research and patch analysis. The concern about patch reverse engineering also reflects AML.T0063 (Discover AI Model Outputs) in a generalised sense: AI is being used to extract actionable intelligence from public artifacts (patches). From an OWASP perspective, the implicit warning against overreliance on patch secrecy as a security control aligns with LLM09 (Overreliance).

Impact Assessment

The policy change affects any software vendor that receives vulnerability reports from Project Zero, which includes the largest technology companies globally. Vendors with slow internal security patch pipelines face the highest risk, as they may be unable to remediate within 30-day windows by 2028. End users of widely deployed software products are indirectly affected — faster disclosure means faster access to patches, but also faster exposure if vendors fail to meet the new timelines. The removal of grace periods eliminates a buffer that has historically protected users during staged rollouts.

Mitigation & Recommendations

  • Invest in patch automation: Vendors should build automated patch generation and testing pipelines capable of delivering fixes within 30 days of a report.
  • Deploy virtual patching and WAF rules: Use runtime mitigations to protect users while full patches are in development, reducing dependence on patch soak time.
  • Assume patches are immediately analysable: Treat every released patch as immediately reversible by an LLM-assisted attacker and prioritise user update adoption accordingly.
  • Engage with Project Zero early: The policy allows vendor-requested early disclosure; build relationships and processes to collaborate proactively.
  • Accelerate update delivery infrastructure: Reduce patch gap through faster auto-update mechanisms, particularly for consumer-facing products.

References

◉ AI THREAT BRIEFING

Stay ahead of the threat.

Twice-weekly digest of critical AI security developments — every story mapped to MITRE ATLAS and OWASP LLM Top 10. Free.

No spam. Unsubscribe anytime.