Thousands of Deceptive Ads on Google Exposed Using Google's Own Free AI Tools
PhD candidate Ritik Roongta and Professor Rachel Greenstadt helped build AdLens, which uses Google's own open-weight Gemma models to find deceptive software ads that slipped through Google's review. Of 24 flagged ads the team reported, Google confirmed only six as violations. "We then reported a sample of them back to Google to see how their moderation actually responds," said Roongta, the study's lead author. "The results were surprising, and they point to some real gaps in how Google handles this."
