The Federal Trade Commission (FTC) is considering a policy that would classify ideological bias in artificial intelligence (AI) systems as an unfair and deceptive practice under Section 5 of the FTC Act. This move could enable the commission to regulate the training data and inputs used by AI algorithms. The FTC's proposed policy statement, released last month, suggests that consumers expect AI systems to provide information free from bias or ideological manipulation.
The proposal has drawn significant criticism from across the political spectrum, with many questioning the FTC's legal authority and the potential for political censorship. Critics argue that the FTC is overstepping its traditional role of consumer protection and stretching Section 5 beyond its intended scope to address a lack of congressional AI regulation.
One major concern is the lack of clarity on how the FTC would define and identify ideological bias in AI systems. While the commission suggests this regulatory authority would supersede state AI laws, specifically mentioning the Colorado AI Act, state lawmakers are already seeking to delay or eliminate bias audits mandated by that law before it takes effect in 2027.
Public comments on the FTC's proposal highlight two primary issues: that it distracts from legitimate questions about the federal government's role in regulating AI deception, and that it could open the door to political censorship of AI model outputs. Some experts argue that while there's a valid debate about bias and accuracy in models and countering disinformation, the FTC's current approach doesn't adequately address these concerns. Instead, it appears to focus on a power struggle with states over AI regulation and disputes over the perceived political leanings of different AI models.
The policy statement's language and references have also raised eyebrows, particularly its repeated mentions of Anthropic, an AI company that has previously clashed with the Trump administration over AI guardrails and military applications. These references are often framed as examples of ideological bias the FTC aims to eliminate. Conversely, the document omits any mention of Elon Musk or xAI's Grok model, despite Musk's public admissions of intervening to shape Grok's outputs on specific topics, such as South African race relations and the term "MechaHitler," to reflect his personal views. Grok is only cited in a footnote referencing an advertisement for it as a "truth-seeking AI companion for unfiltered answers."
Over 300 comments were submitted to the FTC from various organizations, experts, and members of Congress. Most found the proposal ill-defined and susceptible to politically motivated censorship. However, some groups, like the America First Legal Foundation, a conservative non-profit, urged the FTC to adopt the policy fully. They claim that "frontier models" from companies like OpenAI and Anthropic are programmed to prioritize "ideologically liberal and progressive values" as objective truths, and that regulating this falls under the FTC's authority because a "reasonable consumer" would expect an AI advertised for its usefulness and reliability not to skew results with liberal ideological views.
Other conservative groups, such as the R Street Foundation, questioned this "consumer expectations" rationale. They argue that this rationale is typically applied when information is omitted, and since most large language models already include disclosure statements, it would be difficult for the FTC to prove consumers were explicitly deceived about a product's ideological leanings.
Bipartisan concerns were also raised by Representatives Josh Gottheimer (D-N.J.) and Michael Lawler (R-N.Y.), who urged the FTC to exempt civil rights-related work from scrutiny, specifically efforts to prevent models from discriminating based on federally protected characteristics like race, religion, gender, and age. They emphasized that while AI companies should not falsify facts in the name of fairness, they must also prevent discrimination and stereotypes.
The most widespread concern across the political spectrum is that the FTC's proposal could set a precedent, allowing future administrations, including a potential Trump White House, to reshape AI systems to align with their political views.
