Google has paused its Open Source Software Vulnerability Rewards Program after what it describes as a sharp increase in automated reports, most of them invalid. TechCrunch reported the decision on October 4, saying the pause took effect on October 1. The company has promised an update in the first quarter of 2027, rather than announcing a definite reopening date. For researchers using this particular rewards programme, the immediate change is a suspension; for Google, the stated problem is the quality of incoming submissions, not a shortage of reports.
A rewards programme encounters a validation problem
The programme paid researchers for identifying vulnerabilities in Google’s open source software. Its purpose depended on submissions describing genuine security problems. Google’s explanation for the pause points to a breakdown at that first stage: a significant rise in automated submissions, with the vast majority deemed invalid. More incoming reports therefore did not necessarily mean more useful findings for the people reviewing them.
According to TechCrunch, Google announced the change both on X and on the programme’s website. The stated effective date was October 1. The report does not provide a submission count, a rejection percentage or a comparison with earlier reporting volumes. Google’s description establishes the company’s explanation for its decision, but it does not offer enough detail to measure the increase independently.
The scope also matters. This is a pause in the Open Source Software Vulnerability Rewards Program, not a reported closure of every Google bug bounty programme. TechCrunch says participants are being encouraged to consider the company’s other rewards programmes in the meantime. That alternative does not make the suspended programme available again, but it distinguishes this decision from a general abandonment of vulnerability rewards.
AI submissions are the reported pressure, not proof of every failure
TechCrunch frames the problem as a rise in AI submissions. Google’s quoted explanation refers to automated reports and says that most are not valid. Those descriptions connect the decision to the influx of machine-produced material, while stopping short of documenting the origin and accuracy of each submission. The report supplies no breakdown of which tools were used or how Google determined that a report was automated.
TechCrunch also cites Tom’s Hardware as reporting that Google engineers and open source maintainers were overwhelmed by submissions that were invalid or contained hallucinations. That account adds a reported consequence for the reviewers. It remains a secondary account within TechCrunch’s article, rather than a set of published workload measurements or examples that readers can examine individually.
The distinction between a plausible-looking report and a valid vulnerability is central to Google’s stated reason for stopping the programme. A volume increase is not itself a security finding. Nor does Google’s description establish that every AI-assisted submission is wrong. It concerns the predominance of invalid automated reports in the submissions confronting this programme, without providing an assessment of all possible uses of AI in security research.
The next commitment is an update, not a confirmed restart
Google has said it will provide an update during the first quarter of 2027. That is the next stated milestone. Although TechCrunch describes the programme as paused until next year, the announcement reported in the article does not identify a reopening day or promise that ordinary operations will resume immediately when the quarter begins. Treating the update as a guaranteed restart would go beyond the information available.
The source does not set out a replacement submission process, new review criteria or a technical system intended to filter invalid reports. Those omissions leave the response beyond the pause unresolved. The announced action limits this programme for now; it does not explain how Google intends to restore a manageable flow of useful vulnerability findings.
TechCrunch notes that cybersecurity experts had warned the previous year about low-quality AI material threatening bug bounty programmes. Google’s decision gives that concern a specific operational example, based on the company’s own account of its submissions. What remains to be seen is the content of its promised update: whether the programme returns, under what conditions, and how the reported validation problem will be addressed.










