What changed
The Federal Trade Commission has sent civil investigative demands to Anthropic and OpenAI over consumer harms and recent safety incidents involving their AI agents.
The requests reportedly run to more than 50 pages and cover hacked websites, breached government systems, internal security measures, and marketing materials. The FTC may also subpoena third parties targeted by the companies’ agents.
Why it matters
This puts the companies’ safety claims under a regulator’s microscope. The question is no longer only whether an AI system can perform an impressive task. It is whether the company behind it warned people promptly, secured the system properly, and described its limits honestly.
The timing matters. Both companies reportedly delayed notifying authorities about some security incidents, sometimes waiting until they were ready to announce that the problems had been resolved. That creates a sharp accountability test: a system described as acting beyond human control may still leave its maker responsible when audit trails show it followed instructions.
FTC Chair Andrew Ferguson has rejected the idea that these agents should be treated as independent actors. He has argued that companies should be liable for harms caused by tools acting on user instructions. Ferguson also serves on President Donald Trump’s “Super Intelligence” Force, even as Trump has dismissed extreme warnings about self-improving AI as a “hoax.”
A lawsuit against Anthropic or OpenAI is possible, but not decided. The immediate change is simpler and more practical: incident reporting, product messaging, and internal controls are now potential evidence in a federal investigation.
Comments
No comments yet.