The FTC’s New AI Policy: A Return to "Nanny" Regulation or a Necessary Step in Consumer Protection?

The Federal Trade Commission (FTC) is once again drawing parallels to its controversial past, this time venturing into the complex landscape of artificial intelligence with a proposed policy statement that could significantly alter how AI developers operate. The agency’s new stance, detailed in its “Suppression of Accuracy in Artificial Intelligence Systems” proposal, suggests that AI developers may engage in false advertising if they “steer” their AI models’ outputs toward objectives not explicitly understood or desired by users. This move echoes a 1978 FTC proposal to limit television advertising of sugary foods during children’s programming, which was met with sharp criticism and ultimately curtailed by Congress, becoming a historical footnote on regulatory overreach.

The current FTC proposal, announced on July 1, posits that AI companies market their products as helpful and therefore consumers expect maximally accurate responses. Any undisclosed “editorial shaping” of an AI model’s output, according to this theory, constitutes deception. This broad interpretation of deceptive advertising has raised significant questions about the agency’s approach to emerging technologies and its potential impact on innovation and free expression.

Historical Echoes: The "Kidvid" Precedent and Regulatory Overreach

The FTC’s foray into regulating advertising directed at children in the late 1970s serves as a potent historical precedent. In 1978, the agency, tasked with policing unfair or deceptive advertising practices, considered rules to restrict TV commercials for high-sugar foods on programs aimed at young audiences. The Washington Post’s editorial board famously derided the plan as an attempt to “turn the agency into a great national nanny.” Congress, responding to public and industry concerns about overregulation, significantly limited the FTC’s authority in this area, effectively ending the “kidvid” debate for decades. This episode is often cited as a cautionary tale about the potential for regulatory bodies to overstep their bounds in the name of consumer protection.

Now, nearly five decades later, the FTC appears to be adopting a similar "nanny-like" posture, albeit with a novel legal theory applied to the burgeoning field of AI. The agency’s current proposal, unveiled in a proposed policy statement on its website and filed with the Office of the Federal Register, suggests a broad interpretation of deceptive advertising that could ensnare a wide range of AI applications.

The FTC’s Novel Theory: "Steering" as Deception

At the heart of the FTC’s new proposal is the concept of "steering." The agency contends that AI developers "likely" commit false advertising when they direct their models’ outputs toward objectives that users do not anticipate. The underlying assumption is that consumers, drawn to AI by marketing that emphasizes helpfulness, expect unbiased and maximally accurate responses. Therefore, any undisclosed manipulation or prioritization of specific outcomes, beyond pure accuracy, is viewed as a form of deception.

Critics, however, argue that this policy proposal is "in search of a problem," as it fails to identify specific instances of false advertising or genuinely deceived consumers. The FTC’s statement, they contend, lacks concrete examples to support its broad claims, leading to concerns about its practical applicability and potential for unintended consequences.

Legal Challenges and First Amendment Concerns

The FTC’s proposed policy statement faces significant legal hurdles, particularly concerning established First Amendment jurisprudence. The Supreme Court has consistently affirmed that new technologies do not diminish constitutional protections. In Brown v. Entertainment Merchants Association (2011), the Court held that video games, despite being interactive software sold for profit, are entitled to full First Amendment protection, asserting that constitutional guarantees "do not vary" with the emergence of new mediums.

Furthermore, the Supreme Court’s ruling in Moody v. NetChoice (2024) reinforced the principle that online platforms possess protected editorial discretion in curating and presenting expressive content. The FTC’s proposed framework, which equates the design choices underlying large language models with undisclosed "steering," appears to run counter to these precedents. Legal scholars argue that the design and algorithmic choices that shape AI outputs are functionally equivalent to the editorial decisions protected under the First Amendment for video games and social media platforms. In essence, what the FTC labels as "steering" could be legally characterized as protected "editing."

The Disclosure Dilemma: A Double-Edged Sword

The FTC suggests that developers could mitigate liability by "clearly and conspicuously" disclosing that their AI systems prioritize objectives other than pure accuracy. However, the practical implementation of such disclosures presents a significant challenge, as illustrated by the example of "Truthly," an AI chatbot marketed with a specific Catholic bias. Truthly’s slogan, "Every other AI is built to agree with you. Truthly tells you the truth," highlights its claim of providing a religiously informed perspective.

While Truthly affirmatively discloses its Catholic worldview and disclaims impartiality, seemingly aligning with the FTC’s disclosure mandate, it also asserts that its information is filtered "through truth and morality." This creates a potential conflict for consumers, who might find it difficult to reconcile the chatbot’s biased-yet-truthful disclaimers. Under the FTC’s own disclosure standards, such disclaimers could be deemed ineffective if they fail to clearly convey the nature of the AI’s operation. Paradoxically, a religious chatbot, by fulfilling its core function of generating religiously oriented content, could find itself facing false advertising charges.

Broader Implications for Media and Free Expression

The FTC’s proposed legal logic extends beyond AI chatbots, potentially impacting a wide array of media outlets. The principle of freedom of the press, a cornerstone of the First Amendment, could be jeopardized. Any media organization that promises accuracy while exercising editorial judgment, such as The New York Times with its long-standing motto, "All the News That’s Fit to Print," could theoretically face scrutiny under the FTC’s new framework.

Right-leaning media outlets also face potential risks. Newsmax’s promise of delivering "real news," Breitbart’s stated goal to "report the truth – accurately and fairly," and One America News’s branding as "Your Credible Source for National & International News" all hinge on the idea of presenting a particular version of reality. The FTC’s proposal raises the specter of mandatory, cigarette-style warning labels or continuous disclaimers on news programs detailing their specific editorial criteria to avoid investigations.

This approach contrasts sharply with past FTC actions. In 2004, then-FTC Chairman Timothy J. Muris declined to challenge Fox News’s "Fair and Balanced" slogan, stating that such an inquiry would require an evaluation of news content, a task "the First Amendment leaves to the American people, not a government agency." The current proposal appears to reverse this stance, suggesting a government agency’s role in scrutinizing editorial judgment.

A Shifting Regulatory Landscape and Ideological Concerns

The FTC’s current direction represents a departure from previous administrations, particularly under former Commissioner Andrew N. Ferguson. Ferguson had previously championed regulatory humility, emphasizing that enforcement actions should be "guided by the law, not the personal ideology, politics, or novel legal theories of its chairman or commissioners." During his tenure, he rightly supported accountability for AI applications that genuinely deceived consumers, such as DoNotPay’s “robot lawyer” and Workado’s exaggerated AI-detection claims.

Ferguson also dissented from proposed consent orders against generative AI writing tools like Rytr, arguing that the Commission was unduly punishing "a product that helps people speak." He and Commissioner Melissa Holyoak expressed the view that generative AI’s value often lies in its ability to suggest novel ideas, even if they defy user expectations. Ferguson further vacated the Rytr order, condemning enforcement "unsupported by facts or law."

However, the current Commission appears to operate under different principles. The Supreme Court’s ruling in Trump v. Slaughter has been interpreted by some as impacting the FTC’s statutory independence, potentially leading to a Commission more susceptible to partisan influence. Critics point to the current FTC’s proposal, which uses "equity" in scare quotes and criticizes Colorado’s AI law while overlooking similar legislation in Texas and Utah. This selective focus, they argue, reflects an ideological bias rather than a neutral application of consumer protection law.

The administration’s requirement for federally purchased AI models to conform to its "official version of the truth" further fuels concerns about a top-down imposition of a singular worldview. The potential for future administrations to swing the ideological pendulum in the opposite direction raises the specter of ongoing instability and whiplash for both consumers and AI developers, dictated by the prevailing political winds rather than established legal principles.

The Peril of Subjectivity and Stifled Innovation

The FTC’s proposed policy statement is criticized for cloaking subjective assessments in the guise of objective standards. Drawing an analogy to the film Mary Poppins, where the titular character uses a subjective tape measure calibrated with character traits, the FTC’s approach is seen as an attempt to impose a singular worldview under the guise of consumer protection. This, critics argue, risks fostering conformity and stifling the very innovation that is crucial for AI advancement.

The FTC’s assurance that developers can avoid liability through disclosures that "dispel[s] the notion that the system is designed to give the best answer possible" is dismissed as a "piecrust promise," easily made and easily broken. The regulatory uncertainty created by this broad and potentially subjective interpretation of deceptive advertising is seen as a significant threat to the continued development and deployment of AI technologies. Instead of fostering responsible innovation, the FTC’s proposal could inadvertently chill the very advancements it claims to protect, leading to a future where AI development is hampered by an environment of constant legal ambiguity and the potential for politically motivated regulatory action.

The article was written by Keith R. Fentonmiller, who served for over two decades as a senior attorney in the FTC’s Division of Advertising Practices. He is also a published fiction author. The views expressed in his analysis are his own and do not necessarily reflect those of his former employer.