Researchers use Google's free AI to uncover thousands of malicious ads
Researchers found thousands of deceptive ads on Google that bypassed its safety systems. They identified them using Google's free Gemma AI tools, highlighting a gap in detection.
The discovery underscores a growing arms race in online advertising, where malicious actors continuously refine tactics to evade automated filters. Google’s own safety systems are designed to catch deceptive content, but these findings suggest gaps remain. By leveraging Gemma, a freely available AI model, researchers demonstrated a low-cost method for auditing platform defenses—turning the same generative technology that powers many modern tools against the abuse it can enable. This approach highlights how accessible AI can serve as a double-edged sword, both for attackers seeking to craft convincing ads and for defenders seeking to expose them. The scale—thousands of ads—points to a systemic challenge rather than isolated incidents, raising questions about the sufficiency of current review processes.
This story could affect everyday internet users, who may encounter deceptive ads that appear legitimate, risking financial scams or malware. It may also pressure Google and other platforms to strengthen automated detection, potentially leading to stricter ad review policies. Advertisers and small businesses could face more scrutiny, while cybersecurity researchers might gain new, affordable tools for oversight. However, the impact depends on how quickly platforms respond—delays could erode user trust in online advertising ecosystems.