FTC Floats AI Policy Aiming To Ensure That AI Makers Disclose The Truth About Biases In Their LLMs

The U.S. Federal Trade Commission has proposed an AI policy aimed at generative AI and large language models, requiring AI makers to disclose biases that can produce less-than-truthful outputs. The proposal seeks alignment with the FTC Act’s consumer-protection mandate, arguing that people reasonably expect AI to provide accurate information and that undisclosed deviations could be deceptive. While it would not ban bias, it would require companies to tell users when system objectives differ from what users expect. Issued July 1, 2026, the initiative is open for public comment until July 31, 2026. The discussion notes the FTC’s prior focus on false AI advertising and highlights ongoing debate over whether the rules could go too far or are needed to protect consumers. The proposal’s key challenge is defining “truthful” and disclosure methods.