The FTC’s new inquiry should demand evidence from AI companies—and clear limits on government power.
If an AI company wants Americans to trust its products with more consequential tasks, asking how those products fail should not be treated as an attack on progress. Trust has to rest on something sturdier than a polished demonstration and a promise that the engineers are handling it.
Reuters reported September 30 that the Federal Trade Commission had opened an industry-wide probe involving OpenAI, Anthropic and other AI laboratories over potential consumer dangers. A senior FTC official confirmed the investigation to the news agency. The reporting described plans to demand information and testimony about AI-agent security incidents.
That is an investigation, not a finding that any named company broke the law. Reuters said OpenAI, Anthropic and the evaluation organization METR had not immediately responded to requests for comment at publication. The distinction matters, especially when regulators can damage a reputation long before proving a case.
It is nevertheless a reasonable moment to ask what customers are actually being promised, and what evidence supports those promises.
Innovation does not cancel responsibility
A conservative defense of enterprise should begin with voluntary exchange and honest dealing. Buyers need enough information to understand what they are buying. When a seller advertises a capability, the burden should fall on that seller to substantiate it, not on the customer to discover its limits through an expensive mistake.
There is nothing anti-business about that principle. A market becomes less trustworthy when impressive language does more work than reliable performance. Responsible developers also have an interest in competitors being held to the same standards.
For AI, the useful questions are practical. What can a system do without a person approving each step? What limits can the user impose? How are failures recorded? If something goes wrong, can the customer determine what happened and obtain meaningful help?
Those are questions for companies to answer, not accusations that every product fails those tests. An investigation should separate documented problems from dramatic speculation.
Washington needs boundaries, too
The FTC should face scrutiny of its own. Officials ought to explain the conduct they are examining, the authority they are using and the evidence needed to resolve the inquiry. A broad concern about a powerful technology should not become an excuse for indefinite supervision of an entire industry.
Nor should the resulting rules become so elaborate that only the largest companies can afford to comply. Protecting consumers while entrenching incumbents would be a peculiar victory for competition.
The right standard is neither automatic deference to Silicon Valley nor automatic faith in Washington. Companies should be accountable for their representations, and regulators should be accountable for the limits of their power.
Americans can welcome useful technology while refusing to outsource their judgment to its sales pitch. If AI developers want greater trust, showing their work is a better answer than asking the public to stop asking questions.
Claire Whitmore is a conservative writer covering culture, family, education, media, and the rapidly changing expectations placed on women in modern America. Her work examines the intersection between politics and everyday life, particularly debates surrounding motherhood, relationships, career expectations, schools, and popular culture. Whitmore is skeptical of political movements that treat traditional choices as inherently outdated and argues that genuine freedom requires allowing women to define success for themselves. At The Dispatch Room, she brings a measured but unapologetically conservative perspective to some of the country’s most contentious cultural debates.
Discussion
Loading discussion…