Google to automatically publish qualifying bug reports rewarded $10,000 or more
The policy covers submissions from December 1, with waiting periods after fixes and payments. Google wants the reports to help both human researchers and AI agents.
Loading page…
The policy covers submissions from December 1, with waiting periods after fixes and payments. Google wants the reports to help both human researchers and AI agents.
Listen to this story
The policy turns Google’s security bounty reports into a more structured public learning resource as AI agents increase both vulnerability discovery and report volume. Reports submitted from December 1, 2026, qualify only if rewards reach $10,000 and both waiting periods are met: 30 days after a fix and seven days after payment and researcher notification. Special-event and grant reports are excluded. Researchers can choose anonymous publication and propose edits; Google expects disclosures to begin in early 2027.
Google reports 15 times as many duplicate submissions as in 2023, while average report length has tripled as AI agents assist discovery and writing.
Researchers can suggest edits to remove identifiers, cookies, and other sensitive material, and Google will make its own privacy-preserving edits.
Google plans further rule updates, tooling, and additional GitHub mirroring, and urges other bug bounty programs to adopt similar disclosure policies.
Google will automatically publish qualifying security reports submitted to its Google & Alphabet Vulnerability Reward Program from December 1, 2026, giving researchers and AI agents more examples to learn from. The change, announced November 19, covers reports earning at least $10,000, but publication must wait until after the flaw is fixed and the researcher is notified.
The program rewards researchers who report security vulnerabilities. Google has long allowed—and sometimes helped with—coordinated disclosure, in which researchers give the company advance notice before making a flaw public. The new policy adds automatic publication for a defined group of rewarded reports, rather than relying only on researchers to disclose their findings.
Eligibility depends on several conditions working together. A large reward alone does not trigger publication, and December 1 is the submission cutoff—not a promise that reports will appear that day. Google sets out four requirements:
Both waiting periods must be satisfied. A report cannot become eligible simply because the fix is old enough if the payment and notification window has not also elapsed. Qualifying reports will appear at bughunters.google.com/report/reports.
Automatic disclosure does not mean researchers must attach their names to their rewards. Google says some researchers do not want a public connection between their identity and a dollar amount. They will be able to choose anonymous disclosure and propose edits before publication.
Those edits can remove identifiers, cookies and other potentially sensitive material. Google says it will also make its own privacy-preserving edits. The process therefore separates publishing the security finding from exposing personal or sensitive details contained in the original submission.
Google places the policy change in the context of AI-assisted security work. Agents are being used both to find vulnerabilities and to write reports, the company says. Its stated response is to make valid findings more useful as learning material for humans and agents—not just to change how reports reach the public.
The company argues that reviewing valid reports helps security researchers develop ideas for future work. That is the learning opportunity behind automatic disclosure: making rewarded findings available for others to study. Google also urges other bug bounty programs to adopt a similar disclosure policy for all rewarded reports.
There is already a precedent within Google. Its Chrome and ChromeOS reward programs publicly release reported security vulnerabilities after an appropriate interval. The Google & Alphabet policy extends automatic disclosure to its own qualifying reports, with the reward threshold and exceptions described above.
Google describes this as the first in a series of changes, with further rule updates, tooling and additional GitHub mirroring planned. It tells researchers to expect further updates and disclosed reports starting in early 2027.
Loading discussion...
Join the conversation
Explain who should decide when a fixed flaw becomes public.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.