Google Launches AI-Powered Security Scans for Hospitals and Critical Services

Google has launched Scan for Good, a program that uses its Gemini 3.8 Flash Cyber model alongside Wiz's Red Agent pentesting tool to hunt for security exposures in public-facing systems belonging to hospitals, municipalities, public transit operators, and technology providers. The company says the AI systems have already autonomously uncovered critical issues at several organisations during months of behind-the-scenes testing before this week's official global launch.

The AI tools scan publicly accessible websites, APIs, and applications for weaknesses, but only when organisations have explicitly authorised an assessment or under existing bug bounty and vulnerability disclosure programs. Every finding is reviewed by a human security researcher before any disclosure decision is made, according to Wiz's head of offensive security, Gal Nagli. The initiative follows a similar move by OpenAI, which recently announced a billion-dollar credit program to help resource-strapped defenders such as water utilities, community banks, and nonprofits access AI security tools.

Both launches come amid growing scrutiny of AI agents themselves, after reports that AI systems from Google, OpenAI, Anthropic, and Meta have in some cases broken out of their intended boundaries during testing. Google and Wiz argue that keeping humans in the loop for validation and disclosure is central to using these offensive AI tools responsibly.