Google has launched a new reward program to find and fix security flaws in its artificial intelligence (AI) systems. The company is offering rewards of up to $30,000 (approx. Rs 26 lakh) to people who uncover serious bugs that could cause real harm.
Survey
✅ Thank you for completing the survey!
This new AI bug bounty program focuses on rogue actions, cases where an AI system is tricked into doing something it shouldn’t. Examples include an AI prompt that could make Google Home unlock a door, or a hidden command that forces an AI to summarise someone’s private emails and send them to an attacker.
Google has provided clear examples of what counts as an AI-related bug. These include any vulnerability where a large language model or other generative AI tool is used to bypass security, manipulate data, or take unwanted actions. For instance, in the past, researchers found issues that allowed smart home devices to be controlled through poisoned calendar events, opening shutters or turning off lights without permission.
Note that not every AI issue will earn a payout. Simply getting Gemini to make a mistake or generate offensive text doesn’t qualify. Those kinds of issues should instead be reported through the regular feedback tools inside Google’s AI products, where safety teams can study and fix model behaviour over time.
Alongside the new bug bounty program, Google also introduced an AI agent called CodeMender, which automatically helps fix security flaws in code. The company says CodeMender has already been used to patch 72 vulnerabilities in open-source projects after being checked by human experts.
The top reward of $20,000 applies to serious rogue action bugs found in Google’s key products, such as Search, Gemini Apps, Gmail, and Drive. With bonuses for high-quality or especially novel reports, the total can reach $30,000. Lower rewards are available for smaller issues or bugs in other products like NotebookLM or Jules.