R Street Comment on FTC “Policy Statement Addressing AI Accuracy”

Ah, the Federal Trade Commission, that fearless guardian of American commerce, has proposed a Policy Statement Addressing AI Accuracy, and R Street—a think tank about as fond of regulation as a fox is of a kennel—has responded with the regulatory equivalent of a grenade in a punchbowl. They warn that holding AI systems accountable for telling the truth would open a 'Pandora's Box' of 'algorithmic speech policing.' Pandora's Box is a good one, because it implies that without the FTC, the box stays shut and out of it will crawl nothing but truthful, benevolent machines. History has shown us the opposite, of course, but why let a metaphor get in the way of a good deregulatory flourish?

The R Street comment urges the agency to focus 'narrowly' on state violations of interstate commerce rather than on 'how AI companies manipulate system behavior.' Yes, that's right: manipulating system behavior is not a matter for the FTC—only the quaint old interstate-commerce stuff is. So if a chatbot engages in a spot of commerce across state lines while simultaneously inventing a cure for cancer or a libel about your neighbor, the FTC can act on the commerce part and pretend the accuracy part is the speech police's job. In what universe is a chatbot hallucinating a fine—or a lawsuit—not a deceptive practice? Apparently the one where the first principle is keeping the FTC's hands off system behavior at all costs.

The long and short of it is that the FTC's modest attempt to require some degree of truthfulness from AI products has been met with the usual Washington ritual: argue about scope, retreat to procedure, and hope the problem is still there when nobody's looking. It will be—only by then, the Pandora's Box will have been open all along, and the 'manipulated system behavior' will have become the entire economy. But never mind. At least interstate commerce will be scrupulously honest.