Google paused new product vulnerability submissions to its OSS VRP on October 1, 2026, after a sharp rise in automated reports, most of them invalid.
The pause is not a full shutdown. Existing reports, supply-chain bug reports, the Patch Rewards Program, and other Google reward programs remain open.
Curl and Intel faced similar floods, which shows that AI-made reports are straining bug bounty systems across the industry.
Google has paused its open-source bug bounty program, and the move shows how AI is changing security work. The company stopped taking new product vulnerability reports for its Open Source Software Vulnerability Reward Program (OSS VRP) on October 1, 2026. Reviewers faced a flood of AI-generated bug reports, and most were wrong. Google promised an update in the first quarter of 2027.
The OSS VRP rewards researchers who report flaws in open-source projects that Google maintains. It covers the latest versions of projects in public repositories owned by Google organizations. These include Go, Angular, Bazel, Protocol Buffers, and Fuchsia.
Google announced the pause on X and on the program page. It did not close everything. Reports already in the queue still get reviewed, and supply-chain bug reports remain open.
Researchers can also send fixes through the Patch Rewards Program, which pays up to USD 15,000 for high-impact patches. This matters because the pause targets one entry point, not the whole security effort.
Google said the pause followed a sharp rise in automated reports, "the vast majority of which are not valid." Many looked polished but held errors. Some listed wrong triggering conditions. Others described exploit paths that did not exist, or claimed a weak function could be reached when it could not.
Engineers then had to prove each claim false. Reviewers must test every report, so one false claim can cost as much time as a real one. Rewards run from USD 500 to over USD 30,000, depending on the bug and project tier, so the incentive is strong. Generative AI also makes it cheap to send many reports at once. When volume grows faster than review capacity, real bugs wait longer.
Google is not the first to hit this wall. Curl ended its HackerOne bounty at the end of January 2026. Lead developer Daniel Stenberg said the program received seven submissions in one week, and none described a real vulnerability.
He noted that the load on security staff was high. He hoped that ending payouts would remove the push to send poorly researched reports, AI-made or not. Intel also launched a new bounty with no payouts, reportedly to stop a similar flood. Bugcrowd has seen its total submission volume jump. The problem now spans small projects and global tech giants.
Google will use the pause to rework the program. It has already tightened evidence needs for some reports, mainly memory corruption flaws in high-priority projects. The code of conduct also bans low-quality submissions and tells participants to check the output of automated tools.
Google is pointing researchers to its other reward programs while the rebuild happens. Stricter proof may soon become the norm across the industry. These rules will decide whether honest researchers can still earn rewards.
Some experts note that AI can help find real bugs. Stenberg himself noted that strong reports have come from developers using AI as an aid. Malwarebytes points out that an AI-assisted report can be valid, and a poor report is not always AI-made, so drawing a line is hard.
The real issue is unchecked output. Researchers who verify each finding add value, while those who paste raw model output add noise. Good tools can still help when a person confirms the bug before sending it.
Open-source software runs inside phones, cloud systems, and websites. When maintainers spend time disproving fake reports, fixes for real flaws slow down. Many maintainers are small teams or volunteers, so a flood can drain them fast. Security teams at big firms also face more real findings as AI speeds up discovery, which adds to their workload.
Honest researchers also lose, since a paused program cuts off income and credit. For companies, the case shows that bounty systems built before generative AI need new filters. Users feel the effect last, but they feel it: slower patches mean longer exposure to attacks.
Google paused its open-source bug bounty program on October 1, 2026, after AI-generated reports swamped reviewers with mostly invalid claims. The OSS VRP is not gone, since existing reports, supply-chain bugs, and patch rewards continue.
Google plans to rebuild the program and share an update in early 2027. Curl, Intel, and Bugcrowd show that the problem reaches far beyond one company. The lesson is clear: AI can speed up bug hunting, but only careful, verified work keeps open-source security healthy.
1. Why did Google pause its open-source bug bounty program?
Google said a significant rise in automated submissions caused the pause, and the vast majority were not valid. Many AI-made reports held invented exploit paths and wrong assumptions about the code.
2. Has Google shut the program down for good?
No. The pause is temporary. Google plans to rework the program and has promised an update in the first quarter of 2027.
3. Can researchers still report bugs to Google?
Yes, in some ways. Reports already in the queue are still reviewed, and supply-chain bug reports remain open. Researchers can also use the Patch Rewards Program and Google's other reward programs.
4. Are other companies facing the same problem?
Yes. Curl ended its HackerOne bounty in January 2026, and Intel reportedly launched a new bounty with no payouts. Bugcrowd has also seen its submission volume jump.
5. Does Google ban AI tools for bug hunting?
No. The program's code of conduct bans low-quality submissions and asks participants to verify the output of automated tools. A report made with AI help can still be valid if a person confirms the bug.