On Monday, October 6, 2025, Google announced a new bounty program specifically targeting AI‑related vulnerabilities. The program will pay cash rewards for security defects affecting Google Search, Google Drive and other products that embed or rely on generative AI features, with a maximum payout of $30,000 per vulnerability.
According to The Verge, Google frames this initiative as a supplement to its existing Vulnerability Reward Program, with a focus on flaws directly related to AI behavior that could lead to information disclosure or abuse. Concrete examples include attacks that cause a model to expose sensitive data from its context, input constructions that can bypass content and safety policies, or chained vulnerabilities that cause systems to perform unauthorized actions.
The program details emphasize reproducibility and verifiability — researchers must submit reproducible exploitation paths and necessary PoC (proofs of concept) so the security team can reproduce and triage the issue. Reward amounts will be tiered based on technical difficulty, impact scope and exploitability; the quality and completeness of the report will also affect the final payout.
The Verge notes that Google clarified the program's scope and exceptions: not all AI‑related findings are eligible, and cloud services, third‑party integrations or internal systems not in public testing scope are often excluded. Google also reiterated its support for responsible disclosure and encouraged researchers to submit findings through the prescribed process so engineering teams can patch issues promptly.
For the security community and "white hats," the program has two clear attractions. First, a $30,000 cap makes AI‑focused vulnerability research a more compelling career or side opportunity; second, explicitly prioritizing generative AI security steers researchers' attention from traditional software bugs toward model behavior, prompt injection, context exposure and other risks more relevant to current AI applications.
Reports also say Google will consider the visibility of impact during vulnerability assessment: issues that can be triggered at scale or lead to sensitive data leakage generally receive higher ratings and larger rewards. At the same time, Google hopes external researchers will help uncover "chain attacks" — problems that appear low risk in isolation but can cause serious consequences when combined.
For junior bug hunters, this means lower barriers to entry — not only because tools and information are more plentiful, but because vendors are willing to invest more cash incentives in AI security. The Verge interprets the move as a sign that major tech firms are starting to integrate the external security community into their defense loop as they address generative AI risks.
Google's announcement is both an invitation to the technical community and a public commitment: as AI rapidly rolls out into search, productivity and everyday tools, vendors will more actively turn security research into product fixes. For those who wish to participate, the key is to submit clear, reproducible reports and to test within compliance so they can earn rewards while helping build a safer AI ecosystem.