Google Offers Up To $30,000 To Report Bugs In Its AI Products, Check Details
Google defined specific parameters on their blog for what would constitute a bug report in its blog post where it announced the program.
You can earn up to $30,000 from Google if you find and report bugs from their AI offerings.
The tech giant announced on Tuesday its 'Bug Hunters' program which offers base rewards of up to $20,000. They have a bonus of an additional $10,000, if the bug reports are of exceptional quality and include unique observations.
These are for core products such as Gemini, Google Search, Gmail and Drive. Interested parties can use their official website to report these bugs.
But one can't just induce the AI to hallucinate and call up the company to claim their reward.
Google defined specific parameters on their blog for what would constitute a bug report in its blog post where it announced the program. These parameters include 'Rogue Actions' at the very top priority. These are attacks that threaten the users' data and account security leaving it vulnerable to alteration.
The second-most prominent report type was 'Sensitive Data Exfiltration' which are attacks that leak sensitive user information such as address details, bank account details and medical records.
The third-most prioritised report type is 'Phishing Enablement', which involves finding a chink in the AI product's security which lets attackers inject malicious code that would trick users into giving sensitive information.
The fourth-most important report type is 'model theft' which, as the name suggests, involves discovering vulnerabilities that let hackers steal confidential information about how the AI model operates.
Other report types of note include 'Context Manipulation' (to execute Rogue Actions), 'Access Control Bypass', 'Unauthorized Product Usage', and 'Cross-user Denial of Service'.
Along with AI hallucinations, copyright infringement and reports of inappropriate content generated (such as hate speech) will also not count, the firm said. It stated there are already channels of communications built into their AI products to report these problems.
The company invited external researchers over the last two years to find such bugs and disbursed up to $430,000 in rewards.
