Industry

Google Halts Open Source Bug Bounty Program Until 2027 After Wave of Automated AI Vulnerability Submissions

Google paused its Open Source Software Vulnerability Reward Program (OSS VRP) until 2027. Security triagers were overwhelmed by tens of thousands of low-quality, AI-generated vulnerability claims submitted by automated script runners seeking cash rewards.

By Marcus Chen · 2026-10-06 · 10 min read

Google confirmed on October 6, 2026, that it has temporarily suspended its Open Source Software Vulnerability Reward Program (OSS VRP) until early 2027. The program, which pays independent researchers up to $31,337 for discovering security vulnerabilities in open-source projects maintained by Google (such as Chromium, Go, Bazel, and Angular), succumbed to an overwhelming influx of automated AI-generated vulnerability reports.

The suspension highlights how autonomous AI code review tools, when hooked to automated submission forms by financial bounty seekers, can paralyze corporate security operations.

The Numbers: 88.4% Invalid AI Submissions

According to internal telemetry shared by Google's Security Engineering team, the ratio of valid to invalid bug submissions shifted drastically over the past twelve months:

┌────────────────────────────────────────────────────────────────────────┐
│                   Google OSS VRP Submission Ingestion Rates            │
├────────────────────────────────┬───────────────────┬───────────────────┤
│ Quarter                        │ Total Reports     │ Invalid / AI Rate │
├────────────────────────────────┼───────────────────┼───────────────────┤
│ Q3 2024 (Pre-Agentic Surge)    │ 1,240 submissions │ 18.2% invalid     │
│ Q3 2025                        │ 4,890 submissions │ 54.1% invalid     │
│ Q1 2026                        │ 14,200 submissions│ 72.6% invalid     │
│ Q3 2026 (Program Halt)         │ 38,900 submissions│ 88.4% invalid     │
└────────────────────────────────┴───────────────────┴───────────────────┘

Instead of finding real security vulnerabilities, automated agent scripts prompted language models with raw GitHub repositories. When models hallucinated theoretical memory leaks, race conditions, or unvalidated input flaws, scripts immediately submitted the output to Google’s bug bounty portal without human verification.

Impact on Site Reliability and Security Teams

Reviewing each bug bounty submission requires an experienced security engineer to clone the repository, attempt to reproduce the flaw, and write a formal triage response.

The consequences of the AI report surge were severe:

  1. Triaging Delays: Time-to-first-response for genuine, critical CVE disclosures surged from 48 hours to 29 days.
  2. Staff Burnout: Google engineers spent thousands of hours debunking speculative text that contained no executable exploit code.
  3. Open-Source Maintainer Fatigue: External maintainers received dozens of automated pull requests claiming to fix non-existent vulnerabilities, cluttering public GitHub issues.

Requirements for the 2027 Program Reboot

Google emphasized that the suspension is temporary while its engineering team redesigns submission gating:

Legacy 2024–2026 Policy New 2027 Planned Policy
Text-only theoretical descriptions accepted Mandatory executable Docker PoC container
Anonymous or pseudonymous submissions Verified GitHub researcher identity verification
Zero penalty for repeated invalid reports Reputation score penalty; 3 invalid strikes = 1-year ban
Manual human initial triage Automated sandbox compilation verification before human review

Google will continue servicing critical security reports submitted directly via coordinated vulnerability disclosure channels for enterprise infrastructure during the hiatus.