The Federal Trade Commission (FTC) has proposed a policy statement that would apply existing consumer protection laws to undisclosed biases in artificial intelligence models. This move signals a federal effort to regulate AI's potential to steer users toward specific viewpoints, treating deceptive AI practices as a violation of federal law.
The proposal, spearheaded by FTC Chairman Andrew Ferguson, interprets Section 5 of the FTC Act to mean that misrepresentations or omissions are deceptive if they are likely to mislead a reasonable consumer and are material to their decisions. Under this framework, an AI system presented as neutral while systematically promoting particular narratives could be deemed unlawful.
This initiative draws from the Trump administration's AI Action Plan, which emphasized that American AI must be developed to "pursue objective truth rather than social engineering agendas." The plan also included measures to prevent federal procurement of biased AI models and establish a federal standard for model accuracy.
Research indicates that AI models may exhibit undisclosed left-leaning biases, even when trained on factual data. Studies by The Washington Post and MIT's Center for Constructive Communication have found that leading AI models consistently favor liberal arguments on sensitive political topics, often presenting these positions as neutral. These biases are particularly concerning given AI's growing role in how individuals gather information, complete work, and make decisions, including political choices. A New York Times report highlighted voters increasingly using AI chatbots as research tools to help determine candidate support.
The FTC's proposed policy aims to establish a single federal standard for AI models, similar to regulations for other national products like automobiles and pharmaceuticals. This approach seeks to prevent a patchwork of state laws, such as those emerging in New York and California, which could incentivize companies to alter AI outputs to avoid legal liability and potentially penalize accurate responses.
Nicholas Elliot, director of Government Affairs for Innovation Council Action, stated that the American public has a right to know if they are being deceived by AI. The FTC's proposed policy is seen as a critical step toward ensuring AI's responsible development and maintaining American leadership in the AI field.