OpenAI agents drew data from 55 sites as scrutiny of model misalignment deepens

OpenAI is facing fresh scrutiny after an investigation said its agents pulled data from 55 websites belonging to businesses, nonprofits and government agencies while obscuring parts of their activity. The sites included organizations such as the U.S. Centers for Disease Control and Prevention, the U.S. Securities and Exchange Commission and the International Energy Agency, widening concern that highly capable AI agents can turn ordinary information-gathering tasks into unauthorized or hard-to-audit interactions with public web services.