Overview:
-
Google paused new product vulnerability submissions to its OSS VRP on October 1, 2026, following a surge in automated reports, the majority of which lacked validity.
-
This measure is not a permanent shutdown, as existing submissions, supply-chain bug reports, the Patch Rewards Program, and other Google incentive initiatives remain active.
-
Similar deluges of AI-authored reports have impacted Curl and Intel, highlighting industry-wide strain on vulnerability reward systems.
Google’s decision to suspend its open-source bug bounty program highlights the evolving impact of artificial intelligence on security operations. Effective October 1, 2026, the company stopped accepting new product vulnerability reports for its Open Source Software Vulnerability Reward Program (OSS VRP). Reviewers were overwhelmed by a wave of AI-generated bug reports that proved mostly inaccurate. Google has indicated that it will provide an update during the first quarter of 2027.
What Google Paused and What Stays Open
The OSS VRP compensates researchers who identify vulnerabilities within open-source projects supported by Google. The initiative encompasses current versions of code found in public repositories managed by Google organizations, including Go, Angular, Bazel, Protocol Fibers, and Fuchsia.
The company communicated the suspension via X and the official program page. However, operations have not completely ceased. Reports already queued are still undergoing evaluation, and channels for supply-chain bug reports remain open.
Furthermore, researchers can submit fixes to the Patch Rewards Program, which offers up to USD 15,000 for impactful patches. Consequently, the suspension targets a single intake channel rather than Google’s entire security framework.
Why AI-Generated Bug Reports Overwhelmed Reviewers
Google attributed the hiatus to a sharp increase in automated submissions, noting that “the vast majority of which are not valid.” Although many reports appeared well-structured, they contained significant errors, such as incorrect triggering conditions, nonexistent exploit paths, or false claims regarding the accessibility of vulnerable functions.
Security engineers were subsequently forced to manually invalidate each assertion. Because reviewers must evaluate every submission thoroughly, a single erroneous claim demands just as much time as a genuine one. With financial rewards spanning from USD 500 to over USD 30,000 based on severity and project tiers, the monetary incentive is high. Meanwhile, Generative AI enables the rapid and inexpensive production of bulk submissions. When submission volume outstrips review bandwidth, genuine vulnerabilities experience longer delays.
Curl, Intel, and Bugcrowd Face the Same Problem
Google is not alone in encountering this hurdle. Curl terminated its HackerOne bounty initiative at the conclusion of January 2026. Lead developer Daniel Stenberg reported that the project received seven submissions over a single week, none of which outlined a legitimate vulnerability.
Stenberg highlighted the immense strain placed on security personnel, expressing hope that eliminating financial payouts would reduce the incentive to dispatch poorly vetted reports, whether AI-generated or otherwise. Similarly, Intel introduced a new bounty framework without monetary payouts to counter a comparable influx of submissions. Meanwhile, Bugcrowd has experienced a surge in overall submission volume, demonstrating that this challenge affects both niche projects and major technology firms.
How Google Plans to Fix the Program
Google intends to use the suspension period to restructure the program. The company has already increased evidentiary requirements for specific reports, focusing particularly on memory corruption vulnerabilities within high-priority initiatives. Additionally, the updated code of conduct prohibits low-quality submissions and mandates that participants verify the outputs generated by automated tools.
While the overhaul is underway, Google is directing researchers toward its alternative reward schemes. Stricter verification standards could soon become an industry-wide precedent, ultimately determining whether legitimate researchers maintain access to financial incentives.
AI Is Not the Real Villain
Industry specialists point out that artificial intelligence can successfully aid in identifying authentic vulnerabilities. Stenberg observed that valuable reports have indeed originated from developers utilizing AI as a supportive tool. Malwarebytes notes that an AI-assisted report can be entirely valid, just as a flawed report is not exclusively the product of AI, making clear distinctions difficult.
The core issue lies in unverified output. Researchers who rigorously vet individual findings deliver constructive value, whereas those who transmit raw model outputs merely generate noise. Ultimately, automated instruments remain useful provided a human researcher validates the defect prior to submission.
Why This Pause Matters
Open-source infrastructure underpins modern mobile devices, cloud environments, and web applications. When project maintainers expend their resources refuting fraudulent reports, the development of patches for legitimate vulnerabilities slows down. Given that many maintainers operate as small teams or volunteers, a high volume of spurious claims can quickly exhaust them. Concurrently, corporate security divisions face mounting workloads as AI speeds up discovery for real vulnerabilities.
Honest researchers are also penalized, as a halted program eliminates potential revenue and recognition. For enterprises, this situation demonstrates that bounty frameworks established prior to the advent of generative AI require updated protective filters. Ultimately, end users bear the final impact: delayed patching translates to extended vulnerability windows.
Final Thoughts
Google’s temporary suspension of its open-source bug bounty program on October 1, 2026, was triggered by an influx of invalid, AI-generated submissions that overwhelmed reviewers. Nevertheless, the OSS VRP is not permanently shuttered, with queued reports, supply-chain vulnerabilities, and patch incentives remaining operational.
Google intends to revamp the initiative and provide updates in early 2027. Experiences at Curl, Intel, and Bugcrowd illustrate that this obstacle extends well beyond a single entity, reinforcing the lesson that while AI can accelerate vulnerability discovery, maintaining open-source security ultimately depends on careful, verified human oversight.
FAQs
1. Why did Google pause its open-source bug bounty program?
Google pointed to a dramatic increase in automated submissions—the vast majority of which were invalid—as the primary reason. Many AI-constructed reports featured fabricated exploit paths and erroneous assumptions concerning the underlying source code.
2. Has Google shut the program down for good?
No. The interruption is temporary. Google plans to restructure the initiative and has committed to providing an update during the first quarter of 2027.
3. Can researchers still report bugs to Google?
Yes, through specific channels. Reports already present in the queue continue to undergo evaluation, and intake channels for supply-chain vulnerabilities remain operational. Researchers may also utilize the Patch Rewards Program alongside Google’s alternative incentive offerings.
4. Are other companies facing the same problem?
Yes. Curl concluded its HackerOne bounty in January 2026, Intel reportedly deployed a non-payout bounty model, and Bugcrowd has observed a significant elevation in submission volumes.
5. Does Google ban AI tools for bug hunting?
No. The program’s code of conduct simply prohibits low-quality submissions and requires participants to validate automated tool outputs. Reports produced with AI assistance remain acceptable provided a human contributor confirms the authenticity of the bug.
