FTC proposal puts AI-content detector accuracy claims under scrutiny
The proposed Workado agreement highlights the need to examine detector performance claims across content types and AI systems before procurement or governance decisions.
The Federal Trade Commission published a proposed consent agreement concerning Workado, LLC, formerly Content at Scale AI. The agency alleged that Workado’s AI text detector claim, more than 98 percent accuracy across text from generative AI programs including ChatGPT, Claude, and GPT-4, was false or misleading or lacked substantiation. The FTC’s analysis said testing showed around 53 percent accuracy for non-academic AI-generated text.
The proposed order would cover products that detect or purport to detect AI-generated or AI-altered text, images, or video. It would address efficacy representations, requiring them to be supported by competent and reliable evidence, and would require retention of substantiating evidence and public posting of a statement describing that evidence. Other proposed provisions would include customer notice, distribution and acknowledgment requirements, compliance reports, record creation and retention, and compliance monitoring.
The agreement was accepted subject to final Commission approval and placed on the public record for comment. Comments were due by June 2, 2025; the supplied notice did not establish an operative effective date for a final order.