"There is a really interesting debate here about bias and about accuracy in models and whether that’s deceptive or not…about how we counter disinformation that has been absorbed and is now being reflected by LLMs…but this is not addressing that at all," said Leah Siskind, a senior AI fellow at the Foundation for Defense of Democracies.
The FTC’s proposed reach: ideological bias as an unfair or deceptive practice
The Federal Trade Commission is considering treating ideological bias in artificial intelligence systems as an "unfair and deceptive practice" under Section 5 of the FTC Act, according to a proposed policy statement. The commission framed that approach around a consumer expectation that AI systems "will provide them with information free from bias or ideological manipulation," and said defining such bias as unfair or deceptive would potentially let the agency regulate the training data or inputs that power AI algorithms. The document does not fully explain how the FTC would determine when ideological bias exists in a model.
Preemption and the Colorado AI Act
The policy statement also suggests the FTC believes its asserted authority would supersede state AI laws. The statement specifically mentions the Colorado AI Act, which requires models to undergo risk assessments, transparency disclosures and "bias audits" before release. State lawmakers are seeking to delay or eliminate the audits before the Colorado law takes effect in 2027, the record shows.
Audit-ready is a season. It shouldn't be.
Evidence in spreadsheets, controls drifting between audits, frameworks multiplying on flat headcount. Nubivance runs continuous compliance on Rapid7 Cyber GRC - SOC 2, HIPAA, ISO 27001, PCI, CMMC.
End the scrambleExamples cited — Anthropic and the absence of xAI’s Grok
The commission’s sourcing signals a focus on certain companies. Anthropic appears more than half a dozen times in the statement’s footnotes, cited as examples of the kinds of ideological bias the FTC aims to address. By contrast, the document does not name Elon Musk or his xAI-owned Grok model, even though public statements and reporting show Musk has acknowledged intervening when Grok's responses upset him. Those interventions have shaped Grok’s outputs on topics including South African race relations and the term "MechaHitler." Grok appears only in a footnote that cites an advertisement for Grok as "your truth-seeking AI companion for unfiltered answers with advanced capabilities in reasoning, coding, and visual processing."
Criticism from across the political spectrum
The FTC received more than 300 public comments on the proposal, and CyberScoop reviewed dozens of them. Critics on both the left and right raised two common concerns: that the proposal distracts from questions about how the federal government should address AI deception and that it could enable political censorship of model outputs.
- The International Center for Law and Economics wrote the statement "offers little practical guidance about how the Commission will apply its deception authority to AI" and warned the commission's focus "suggests that the Commission’s concerns extend beyond factual misrepresentations in marketing to speech that may receive the highest degree of First Amendment protection."
- The America First Legal Foundation urged the FTC to adopt the policy "in full," arguing that frontier models from OpenAI and Anthropic "have been programmed to prioritize ideologically liberal and progressive values as though they are objective, neutral positions rooted in truth." Emily Percival, senior counsel for America First Legal, wrote that a "reasonable consumer" would not expect a model to "adopt overwhelmingly liberal positions."
- Conservative commenters did not speak with one voice: the R Street Foundation's Spence Purnell and Adam Thierer argued the consumer expectations rationale "is typically used in cases where there is an omission of information that should have existed" and noted that existing LLM disclosure statements make it unlikely the FTC could prove consumers were deceived.
- David Inserra, Jennifer Huddleston and Juan Londoño of the Cato Institute said the FTC was conflating ideological bias and factual deception, calling the agency's comparison of judging AI models to the way it evaluates dietary-supplement claims "an absurd comparison."
- Reps. Josh Gottheimer (D‑N.J.) and Michael Lawler (R‑N.Y.) urged the FTC to carve civil‑rights-related work—preventing models from discriminating based on race, religion, gender, age and other protected characteristics—out of its scrutiny, while noting companies must both avoid falsifying facts and prevent discrimination and stereotypes.
What this means for AI companies, state lawmakers, and consumers
- AI companies: Face the prospect of federal enforcement over model training and inputs if the FTC finalizes a policy classifying ideological bias as deceptive; some firms are cited explicitly in the document (Anthropic), while others mentioned in reporting (xAI/Grok) are absent from the policy text.
- State lawmakers: Those behind laws like the Colorado AI Act may confront a federal claim of preemption over state-mandated risk assessments, transparency disclosures and bias audits that are scheduled to take effect in 2027.
- Consumers: Will be central to the FTC's legal theory—whether a "reasonable consumer" expects ideologically neutral outputs—but commenters dispute whether that consumer-expectation test applies cleanly to accuracy, ideological balance, or model design choices.
The record in public comments and the policy statement frames this as a tug-of-war over legal tools and jurisdiction: the FTC is staking a claim to police ideological bias under consumer-protection law, while numerous critics question both the legal footing and the practical implications, warning of censorship and constitutional entanglements. The commission has gathered hundreds of comments, but how it will define ideological bias and measure it in practice remains unspecified in the proposal.
