Google Suspends Open-Source Bug Bounty After AI-Generated Spam…

On October 1, 2026, Google halted product vulnerability submissions for its Open Source Software Vulnerability Reward Program (OSS VRP), citing a surge in automated, AI-generated reports that were overwhelmingly invalid. Supply-chain reports remain open, while product submissions await an architectural overhaul promised for Q1 2027.

{"main_topic":"cybersecurity","topics":["cybersecurity","bigtech","ai","opensource","vulnerabilita"]}

On October 1, 2026, Google blocked product vulnerability submissions for its Open Source Software Vulnerability Reward Program (OSS VRP). The cause: an influx of automated, AI-generated submissions, the vast majority of which were invalid. The program — one of the primary disclosure channels for independent researchers working on open-source projects — remains partially operational only for supply-chain reports, while product submissions await an architectural overhaul with an update promised for Q1 2027.

Key Takeaways

Google suspended product vulnerability submissions for the OSS VRP effective October 1, 2026, per an official announcement posted on X.
The stated reason is a "significant rise in automated submissions, the vast majority of which are not valid."
Supply-chain reports remain open; researchers can redirect to the Google Cloud VRP or the Patch Rewards Program.
The phenomenon is not isolated: Linux maintainers had already reported being "completely overwhelmed" by roughly 2,000 bogus CVEs per release, and Intel recently froze its own bug bounty.

"Why is this happening? This pause is due to a significant rise in automated submissions, the vast majority of which are not valid. We will continue to reformat and work on this aspect of the OSS VRP and commit to giving an update in Q1 2027." — Google, via post on X (as reported by Times of India)

The Block Structure: What Stays Open and What Doesn't

The suspension is not total. Google spelled out three operational exceptions. Supply-chain reports — those concerning the software dependency supply chain — continue to be accepted within the OSS VRP. Product vulnerabilities already logged before October 1, 2026 retain their triage workflow. Additionally, certain vulnerability types in Google Cloud repositories can still route through the Google Cloud VRP, a separate program unaffected by the pause.

The source does not specify the selection criteria for repositories eligible under the Cloud VRP nor the percentage of researchers who can practically migrate their workflows. Google has pointed to the Patch Rewards Program as an alternative route, without detailing differences in scope or payout compared to the original OSS VRP.

The Collapse Mechanism: Zero Marginal Cost Versus Finite Capacity

The technical core of this episode is economic more than purely computational. Large language models have driven the marginal cost of generating formally plausible vulnerability reports to zero: correct technical language, references to real components, disclosure structure aligned with industry standards. The barrier to entry for a malicious actor or spam operator is no longer reverse-engineering expertise, but prompt-engineering and API automation skill.

Human triage does not scale linearly. A typical bug bounty program spends person-hours verifying each submission: reproducing the exploit, analyzing impact, assessing severity, communicating with the maintainer. When the ratio of valid to invalid reports shifts drastically — Google characterizes it as "the vast majority" invalid — the verification system becomes the structural bottleneck. The suspension is therefore an emergency measure to preserve the program's residual functionality, not a response to a security threat in the classic sense.

The dossier does not specify whether Google detected recurring patterns in the automated submissions: particular LLM models, common prompt structures, or recurring network metadata. Without these data, it remains an open question whether the phenomenon is orchestrated by specific actors or diffuse as the aggregated behavior of individual operators.

Industry Context: Linux and Intel as System Signals

Google's decision fits a broader pattern documented by the same sources. Linux kernel maintainers had already publicly stated they were "completely overwhelmed" by fraudulent CVE reports, with AI-driven vulnerability hunters pushing the count to roughly 2,000 CVEs per release — an order of magnitude exceeding the verification capacity of any open-source security team.

Intel, for its part, recently froze its own bug bounty, which offered payouts up to $100,000 per vulnerability. Editorial sources link this suspension to the same AI-spam bottlenecks, but with necessary caution: Intel has not officially confirmed this motivation. The temporal and sectoral correlation is solid; the documented causality is not.

The convergence of three cases across distinct ecosystems — Google OSS, Linux kernel, Intel hardware — suggests the problem is systemic for disclosure infrastructure, not confined to a single organization or program model. Bug bounty programs were designed in the pre-LLM era, assuming human reporting and natural opportunity costs that filtered noise. The removal of those filters demands an architectural renegotiation.

What to Do Now

Researchers with product reports ready must check the log date: only submissions logged before October 1, 2026 retain an active workflow in the OSS VRP.
Anyone working on vulnerabilities in Google Cloud repositories must confirm eligibility for the Cloud VRP before proceeding, since not all repositories fall under the exception.
For non-Google open-source projects, evaluate the Patch Rewards Program as an alternative, paying attention to differences in scope and eligibility criteria versus the original OSS VRP.
Maintainers of third-party open-source projects should monitor their own disclosure pipelines: manual verification of suspicious CVEs generates security technical debt that delays real patches.

The Promised Overhaul and the Dossier's Limits

Google has committed to providing an update in Q1 2027, without pre-committing to the nature of the reform. The dossier does not document whether the company is evaluating: researcher authentication requirements, AI-generated content detection tools, submission quotas, or modified payout models to disincentivize volume. Also unknown are the financial extent of the damage — wasted triage hours, lost opportunities for legitimate researchers — and the impact on maintainers of open-source projects covered by the OSS VRP, who temporarily lose a structured disclosure channel.

The source also does not specify whether other Google programs — Chrome VRP, Android VRP — face similar dynamics, or whether architectural defenses under development for the OSS VRP will be extensible.

The Read: Toward a New Verification Model

The episode highlights a structural contradiction in contemporary security. Bug bounty programs democratized vulnerability research by lowering the reporting cost for legitimate researchers; generative AI has lowered that same cost for spam operators, nullifying the selective advantage. The result is an informational tragedy of the anticommons: when the scarce resource is triage capacity, overload by a free-access actor penalizes the entire community.

The solution likely lies not in a single filter technology, but in an economic redesign of the program: mechanisms that reconstitute differentiated opportunity costs, making quality reporting worthwhile and mass reporting costly. Q1 2027 will be the test of whether Google — and the industry — have found that lever.

Information verified against cited sources and current as of publication.

Sources

Sources and references

Get DeafLetter