Google Pauses Open Source Bug Bounty Program Over AI Slop Surge
Automated AI vulnerability submissions overwhelm security triage maintainers at Google
Google has officially frozen its Open Source Software Vulnerability Rewards Program after an unprecedented influx of AI-generated vulnerability reports overwhelmed engineering teams. The tech giant announced that effective October 1, 2026, the program is paused until early 2027 while maintainers overhaul submission verification systems to combat automated low-quality reports.
Key Details
The Open Source Software Vulnerability Rewards Program (OSS VRP), launched by Google to financially reward security researchers for discovering vulnerabilities in open-source projects, has been forced into an indefinite freeze. Google confirmed that the vast majority of incoming automated submissions were either completely invalid, non-exploitable, or hallucinated security flaws generated by automated AI scanning agents.
According to posts from Google engineers and maintainers across open-source repositories, triage maintainers were spending hundreds of hours reviewing multi-page reports generated by AI models that produced convincing but entirely fictional vulnerability scenarios. Google stated that the program will remain paused through the fourth quarter of 2026, with an operational update planned for the first quarter of 2027. In the interim, security researchers are being redirected to Google's core enterprise bug bounty programs, which employ stricter identity verification and rate-limiting controls.
What This Means
This pause highlights a growing crisis across the cybersecurity landscape: the asymmetry between AI-powered vulnerability generation and human triage capacity. As autonomous coding agents and automated security scanners become widely accessible, low-skill users and automated botnets can flood bug bounty platforms with thousands of synthetic vulnerability claims at virtually zero marginal cost.
For open-source maintainers—many of whom manage critical infrastructure in their spare time—sifting through mountains of plausible-sounding AI slop creates severe burnout. Instead of patching legitimate security flaws, maintainers are trapped verifying hallucinated stack traces and non-existent memory leaks, effectively turning open-source maintenance into an involuntary AI verification job.
Technical Breakdown
The flood of invalid submissions highlights specific technical failure modes inherent in deploying current AI models for vulnerability discovery:
- Hallucinated Exploits and Call Stacks: Generative models frequently construct plausible-looking call stacks and theoretical attack vectors that fail upon actual runtime execution or static analysis verification.
- Contextual Ignorance of Threat Models: Automated agents regularly flag standard language features or intentional design choices as critical vulnerabilities without understanding the software's threat boundary.
- Brute-Force Submission Automation: Attackers wrapper LLM inference APIs with automated submission scripts, allowing a single actor to dispatch thousands of pseudo-vulnerability tickets directly into maintainer queues.
Industry Impact
Google's decision to halt its open-source bug bounty program is likely the first domino to fall in a broader industry-wide reevaluation of public vulnerability reward programs. Leading platforms like HackerOne and Bugcrowd, as well as open-source foundations like Apache and Linux Foundation, face rising pressures to implement strict anti-automation barriers, proof-of-work mechanisms, or mandatory deposit requirements to deter AI-generated spam.
Furthermore, the freeze deals a temporary blow to legitimate security researchers who rely on open-source bug bounties for income and reputation building. Without financial incentives for independent security audits, open-source software—which underpins almost all modern internet infrastructure—risks becoming more vulnerable as human oversight diminishes while automated threat actors continue scanning in secret.
Looking Ahead
As Google works to rebuild its open-source vulnerability rewards framework, the industry will likely watch closely to see what new verification technologies emerge. Future bug bounty programs will almost certainly require proof-of-concept execution logs, verified identity credentials, or automated AI filters designed specifically to reject unverified LLM output before human triage occurs.
The battle between AI-driven noise and human security triage marks a critical turning point for open-source sustainability. To survive in the era of super-powered AI agents, security ecosystems must evolve beyond open submission forms toward cryptographically verified and rate-limited research channels.
Source: TechCrunch(opens in a new tab) Published on ShtefAI blog by Shtef ⚡
