The bug bounty inbox has always been a weird little swamp: real exploits, enthusiastic misunderstandings, duplicate reports, and the occasional screenshot with the forensic value of a potato. Google just put up a temporary fence around part of that swamp. Not because researchers stopped finding bugs, but because machines made it too cheap to submit noise at industrial scale.
For once, this is not a breach with a laminated apology statement. It is a process failure story, which is somehow more useful and less cinematic. The lesson is blunt: when AI makes report generation cheap, the expensive part becomes proving that anything in the pile is real.
What Happened: A Bounty Queue Hit the Spam Wall
According to dev.ua, Google suspended reports under its Open Source Vulnerability Rewards Program starting October 1, 2026, after an abnormal surge in AI generated reports, most of which were invalid. BleepingComputer reported the same pause for OSS VRP submissions and said the program covers Google maintained open source projects including Golang, Angular, Bazel, Protocol Buffers, and Fuchsia, along with critical third party dependencies and repository settings. Google launched OSS VRP in August 2022, according to BleepingComputer, with rewards ranging from $100 to $31,337 for security flaws with significant software supply chain impact.
@title How AI spam jams triage
@source BleepingComputer
@source Neowin
AI reports
│
▼
Mostly invalid
│
▼
Human review
│
├─ OSS product pause
├─ Supply chain continues
├─ Cloud VRP continues
└─ Patch Rewards continues
@caption Google paused OSS product reports while other reporting paths remain open.
The key quote, via BleepingComputer, is the sort of calm corporate sentence that hides a triage room full of people staring into the fluorescent abyss: "We are temporarily no longer accepting OSS VRP product vulnerability submissions. This does not impact OSS VRP supply chain reports, or any outstanding reports." Google also said, according to the same report, "This pause is due to a significant rise in automated submissions, the vast majority of which are not valid." Translation: the scanner found a shadow, wrote a novella, and a human still had to check whether the dragon existed.
The Blast Radius: Narrower Than The Headline Sounds
Neowin reported that the freeze affects open source bug reporting until the first quarter of 2027, but only for product vulnerability submissions under the OSS VRP framework. That distinction matters, because vulnerability disclosure programs are not one giant red button labeled panic. BleepingComputer reported that OSS VRP supply chain reports and outstanding reports are not affected, and that researchers can still use Google Patch Rewards Program submissions for open source fixes, with bounties of up to $15,000 for high impact fixes.
BleepingComputer also reported that vulnerabilities in Google Cloud open source repositories affecting Cloud products can still go through Cloud VRP. So the useful reader takeaway is not that Google has walked away from open source security. It is that one intake lane got clogged so badly that Google temporarily closed it while leaving other lanes open, like a hospital redirecting non urgent cases because the waiting room is full of AI generated charts diagnosing everyone with lighthouse fever.
The Economics: AI Made Reports Cheap, Validation Stayed Expensive
Neowin reported that maintainers were being overwhelmed by thousands of false reports and that every report had to be checked by a human, leading to burnout. That is the real story. AI did not magically create verified vulnerabilities at the same rate it created plausible sounding submissions, and plausible is a dangerous word in security because it is one costume change away from wasting a week.
Bug bounty programs used to assume a certain amount of friction: finding a bug, reproducing it, writing a report, and explaining impact took effort. AI lowers the cost of the paperwork and sometimes the first pass discovery, but it does not automatically lower the cost of validation. The bottleneck moves from discovery volume to signal quality, and if your intake process pays the same attention tax on nonsense as it does on a working exploit, congratulations, you built a denial of service vulnerability in your own triage queue.
What It Actually Means For You
For builders and security leaders, the Google pause reported by BleepingComputer and dev.ua is a useful case study in designing disclosure programs for an AI saturated world. Intake forms need to reward reproducibility, proof of impact, minimal working examples, and clear affected components, not just confident prose. Programs may also need sharper scopes, pre submission validation checks, researcher reputation weighting, and faster rejection paths for reports that cannot show a real security boundary being crossed.
For researchers, the message is equally practical: AI assistance is not the sin, low signal submission is. If a tool helps you inspect code, great, welcome to modern work. But if the final report cannot explain the vulnerable path, the impact, and the steps to reproduce, it is not research, it is confetti with a CVE shaped hat.
The next thing to watch is how Google readjusts OSS VRP before the pause lifts, which BleepingComputer said Google is working on. If Google finds a better filter, other maintainers will copy it, because nobody wants their bug bounty desk turned into an AI compost heap. The internet is not saved by more reports alone; it is saved by validated, actionable reports that let humans fix the right thing before the wrong people notice.