This obscure FTC proposal exposes the widening fight over what AI can say

More than 300 public comments on the Federal Trade Commission's draft policy targeting AI-generated speech reveal deep concern over government policing of AI content and geopolitical stakes involving China. The policy, which targets speech 'distorted by undisclosed ideological objectives' and 'so-called equity', was broadly panned by groups across the ideological spectrum, with many arguing it gives the FTC a lever to control AI speech. A minority of comments focused on China as a threat, justification for light-touch regulation, or an accusation of censorship, with groups like the Open Markets Institute comparing the proposal to China's censorship of AI models.

AI https://www.fastcompany.com/section/artificial-intelligence companies are increasingly being pulled into a fight over who gets to decide what their systems can and can’t say. And a little-noticed Federal Trade Commission FTC proposal is showing just how sprawling that fight has become. More than 300 public comments on the FTC’s draft policy https://www.ftc.gov/policy/public-comments/policy-statement-concerning-suppression-accuracy-artificial-intelligence-systems reveal deep concern about the government’s role in policing AI-generated speech, and about the geopolitical stakes of doing so. The policy targets AI-generated speech that is “distorted by undisclosed ideological objectives” and “so-called ‘equity’”. But the comments reveal another force shaping the debate: anxiety over China. Many influential industry, legal, and advocacy groups argued that the FTC is setting up its own lever to control what AI services say. The policy was broadly panned by groups across the ideological spectrum. But a striking minority of the comments were preoccupied with China and its impact on American technology policy. China appears as a threat, as a justification for light-touch regulation for competitive reasons, as a foil for U.S. values, and as an accusation that the administration is pursuing censorship. The Center for Media and Digital Governance at the anti-monopoly group Open Markets Institute argued in its response to the FTC draft that content controls can amount to an exploitation of power concentration. It compared the proposal to China’s “censorship of its AI models which systematically refuse to engage with politically sensitive subjects including Tiananmen Square, Taiwan, and criticism of the Communist Party.” The national security think tank Foundation for Defense of Democracies argued, conversely, that content oversight and transparency are necessary to combat bias inherent to Chinese models that are gaining use even in the United States. Innovation Council Action, a group linked to Trump administration technology advisors, argued that the proposal was necessary to help preempt state-level AI policies and thereby stay competitive with China. Those arguments could prove consequential. Ironically, fears about the world’s leading practitioner of content control could help give staying power to a policy that critics see as a First Amendment threat. The FTC already had an eye on China, noting in its draft that “geopolitical rivals are investing heavily in this sphere, hoping to inject their own companies and values into the marketplace.” AI services are deeply global, and China and the United States are thinking through many of the same problems, so it’s no surprise there would be some interaction. Still, content control is among China’s top AI governance priorities—and perhaps the greatest contrast to an idealized U.S. approach, which emphasizes First Amendment protections, disclaims censorship, and embraces choices by market actors. In Xi Jinping’s July speech at the World AI Conference in Shanghai, he called for AI that is “secure and controllable,” for AI that does not “undermine … the uniqueness of cultures of different countries,” and for preventing “malicious use”—concepts that have all been operationalized in China to support political censorship. China has refined and debated its content control systems at several points. Measures released in 2023 direct AI services to protect the “nation’s image” and uphold “socialist core values.” China’s 2025 AI Safety Governance Framework 2.0 highlighted “information content risks.” The global reach of Chinese models means that all of these policies and directives have real effects for international users. Given the contrast, it would be a stretch to suggest that the FTC’s apparent attempt to steer AI content is equivalent to China’s broad ranging content controls. But policymakers should take heed that they are asking private individuals and companies to steer speech—an area where AI services have a clear technical capability. Instead, policy answers to the influence of Chinese models must take a different form. Is China an appropriate counterpoint? As a former diplomat, the question delights me. I’ve long been fascinated by how what’s made in America is shaped by our beliefs, fears, and hopes about what is happening around the world. The U.S. AI policy conversation is animated by China in countless ways. Outpacing, countering, and exposing China are favorite justifications, from industrial policy for chips and data centers to the politics of what AI models say and do. Content control should remain a divergence between U.S. and Chinese AI offerings, first as a question of values and goals. A wide range of models offering different styles and perspectives is healthy. The threat of FTC enforcement, used to steer speech, takes us in the wrong direction.