Google Pauses Bug Bounty Over AI-Generated Spam
Google has suspended submissions to its Open Source Software Vulnerability Rewards Program after being flooded with AI-generated reports, the latest and largest program to buckle under automated low-quality vulnerability submissions.
Key facts
What happened: Google temporarily stopped accepting OSS VRP product vulnerability submissions
The reason: A significant rise in automated submissions, the vast majority of them invalid, per Google
What's unaffected: Supply chain reports, outstanding reports, and submissions made before October 1, 2026
Alternatives: The Google Patch Rewards Program (up to $15,000 for high-impact fixes) and Cloud VRP remain open
The timeline: Google says it will provide an update in Q1 2027
Not the first: curl ended its bug bounty in January for the same reason; Intel removed financial rewards in mid-September
What is a bug bounty, and why does this matter?
A bug bounty program pays outside security researchers to find and responsibly report vulnerabilities rather than selling or exploiting them. It is one of the more effective ideas in security economics — it makes defense pay.
Google launched the OSS VRP in August 2022, with rewards from $100 to $31,337, covering open-source projects it maintains including Golang, Angular, Bazel, Protocol Buffers and Fuchsia, plus critical third-party dependencies and repository configuration.
Since starting its first program in 2010, Google has paid security researchers over $81.6 million. In 2025 it awarded a record $17.1 million to more than 700 researchers, up 40% from $12 million the prior year.
Why are AI-generated reports a problem?
Because triage costs real time regardless of whether a report is valid.
Every submission has to be read and assessed by a human. An AI can generate a plausible-sounding vulnerability report in seconds; determining that it is wrong can take an engineer an hour. When submissions scale and triage does not, the program drowns.
Google's own language is blunt: the pause is due to a significant rise in automated submissions, the vast majority of which are not valid.
Is this an industry-wide pattern?
It is becoming one. In January, the maintainer of curl — the command-line tool and library embedded in an enormous share of internet-connected software — ended its HackerOne bug bounty after being overwhelmed by what he described as AI slop reports.
In mid-September, Intel removed all financial rewards for flaws reported through its Intigriti program, without publicly explaining the decision.
Microsoft warned in May that AI tools surface far more vulnerabilities, that the pace and breadth of vulnerability discovery would increase across the industry, and that this can raise operational demands. Last month it patched a record 966 flaws in a single Patch Tuesday, including two actively exploited zero-days.
What's at stake
Bug bounties are a defensive channel. If they close, researchers who would have reported responsibly face fewer legitimate outlets — while the markets that buy vulnerabilities for offensive use have no triage problem, because they only pay for what works.
The tension is real and unresolved: the same automation flooding these programs is also genuinely finding more vulnerabilities. Google says it is reworking the OSS VRP to handle automated submissions and will report back in early 2027.
For businesses, the practical read is narrower. Expect the volume of disclosed vulnerabilities to keep climbing, and expect patch cycles to get more crowded, not less.