aiqbench analysis
Google is applying its Gemini models, with human security researchers, to find public-facing exposures in public services, critical infrastructure, and nonprofits, and it is working with CISA and Google DeepMind on the effort. The announcement offers no independent metrics beyond "hundreds" of fixed exposures, so the results are self-reported and hard to assess. For AI readers, the notable point is the dual-use framing: the same capabilities that help defenders find flaws could help attackers.
Scan for Good: Using AI to discover and fix high-priority exposures across public services and critical infrastructure

New initiative partners with under-resourced organizations to uncover, remediate exploitable risk at scale.
September 30, 2026 Update: Our Scan for Good initiative now uses Gemini 4 Argon, alongside Gemini 3.8 Flash Cyber. As offensive AI capabilities improve, empowering defenders has become a global imperative. The damage cyberattacks can inflict upon under-resourced organizations and the communities that depend on them is an issue that affects us all. It is in this spirit that we are launching Scan for Good, an initiative that uses AI - coupled with human security researchers - to uncover public exposures and complex attack paths across public services, critical infrastructure, and nonprofits. Scan for Good helps defenders secure vital systems before malicious actors can exploit them. Early results prove its impact. Our efforts spent validating findings and collaborating with the affected organizations on remediation have uncovered hundreds of public exposures that were fixed as a result of Scan for Good. Later in this post, we will share some of these stories. We would like to thank the team at Google DeepMind for partnering with us to ensure the success of Scan for Good. We’d also like to thank the Cybersecurity and Infrastructure Security Agency (CISA) for engaging with us on the project to offer their collaboration and guidance. At a…
Excerpt shown under the publisher's RSS terms · 200 words · full text and figures are on the publisher's site