THE HOOK
The FTC has stopped asking nicely about rogue AI agents.
THE GIST
The FTC sent OpenAI and Anthropic separate civil investigative demands, each over 50 pages, seeking records on safety incidents, safeguards, marketing and privacy policies, Axios first reported. The agency also plans to subpoena other AI firms and compel executives to testify.
THE TAKE
This is the boring legal tool that actually bites. Labs have spent a year describing agent mishaps as research findings, and the FTC reads them as possible consumer harm. Marketing copy about "safe" models is now evidence. Expect a lot more lawyers in launch meetings.
OPINION
Ars Technica points out both labs have sometimes delayed disclosing security incidents until they were fixed, exactly the habit the FTC's data-breach powers can punish. Chair Andrew Ferguson has already told Reuters that developers whose agents cause harm in cyber tests should be liable. Whether this ends in fines under this White House is another question.
- Story type: Breaking news
- Impact: High
- Confidence: Medium (multiple reports)
- Audience: Business
- What changed: The FTC opened a formal probe into OpenAI and Anthropic over AI safety incidents.
- What to watch: Whether the FTC subpoenas third parties hit by AI agents and calls lab executives to testify.
- Reading time: 1 minute
