AI Spam Forces Google to Halt Open Source Bug Bounty

A surge of AI-generated vulnerability reports has forced Google to suspend its open source bug bounty program, halting payouts to security researchers who find flaws in the company’s publicly available software projects.

The pause on the Open Source Software Vulnerability Rewards Program, known as OSS VRP, took effect Oct. 1 and will remain in place through at least early 2027. Google disclosed the suspension on X, attributing the decision directly to a flood of automated submissions that overwhelmed the review process.

According to the company, the vast majority of these AI-generated reports were invalid. Security teams tasked with triaging incoming vulnerabilities found themselves sorting through high volumes of machine-produced noise rather than actionable intelligence, effectively degrading the program’s capacity to identify genuine threats.

Launched in August 2022, OSS VRP was designed to crowdsource the discovery of critical security flaws not only within Google’s own open source repositories but also across third-party dependencies those projects rely upon. Payouts for severe vulnerabilities could exceed $31,000. The financial incentive structure, combined with the public availability of the source code, made the program an ideal target for automated scraping tools powered by large language models.

Valid bug reports submitted before the Oct. 1 cutoff remain eligible for payment. Bounties tied to supply chain vulnerabilities are also unaffected by the freeze, preserving a critical layer of defense for downstream software ecosystems that depend on compromised libraries.

The collapse of manual triage under the weight of generative AI is becoming a systemic issue across major software development communities. Linux kernel creator Linus Torvalds previously moved to restrict AI-assisted patch submissions after automated tools began flooding the project’s mailing lists with low-quality code changes, mirroring the exact bottleneck now paralyzing vulnerability reporting pipelines.

When organizations offer bounties based on code that is freely accessible, automated agents can scan, analyze, and submit findings at a scale no human researcher can match. Because current AI models frequently hallucinate plausible-sounding but technically incorrect code behaviors, the resulting submissions often mimic the format of legitimate security research while lacking actual exploitability. This forces engineering teams to spend disproportionate resources disproving false positives.

Google plans to outline the structural changes required to restart the program during the first quarter of 2027. Any revised framework will likely need to implement strict verification mechanisms or submission caps to filter out automated traffic without discouraging independent security researchers.

The company continues to operate its primary Vulnerability Reward Program for proprietary products like Chrome, Android, and its cloud infrastructure, which remain active and separate from the open source initiative.