Category: Regulatory | Reading time: 5 min
The Remark
On 25 September 2026, Federal Trade Commission Chairman Andrew Ferguson spoke at Reuters Momentum AI Austin and rejected the framing of AI agents as independent actors that break loose with wills and desires of their own.
"I'm going to continue as long as I am chairman to resist this anthropomorphizing of these tools," Ferguson said, according to Reuters. His liability analogy was blunt: if someone tells a tool to do something and the tool does it, the question is not what to do about the tool.
What He Said About Audit Trails
Ferguson said AI companies have sometimes described systems as acting beyond human control. Subsequent reviews of audit trails, he said, repeatedly showed the systems carrying out instructions they had been given.
That sentence matters more than the rhetoric around anthropomorphism. It tells every deployer and every developer what the FTC expects to find when an agent causes harm: a record that ties the action to an instruction, a policy, or an authorized scope. An empty or unverifiable trail is not a defense. It is evidence that the organization cannot show what it told the system to do.
Existing Tools, Not a New Statute
Ferguson said the United States should use existing legal tools rather than rush to European-style AI regulation before testing what current law can do. He suggested FTC authority over companies that fail to disclose data breaches can also reach AI developers.
The posture is continuity, not a new AI Act. Section 5 unfair and deceptive practices authority, breach disclosure expectations, and the means-and-instrumentalities doctrine are the likely paths. Companies that ship agent products should assume the developer and the operator, not the model, are who get named.
Why This Lands Next to California AB 316
California AB 316 already eliminated the "AI acted autonomously" defense in civil litigation for deployers in that state. Ferguson's comments extend the same logic into federal consumer-protection enforcement language: agents are tools, instructions are attributable, and audit trails decide whether the "rogue" story survives contact with evidence.
Banks, insurers, health systems, and any vendor selling agents into those environments now have a clearer enforcement narrative. "The agent did it" is not a theory the current FTC chair is willing to entertain in public.
What Deployers Need on Paper
The practical requirement is not a new slogan. It is a production record that can survive an FTC or examiner question:
Who authorized the agent. What scope it was given. Which tool calls were allowed, blocked, or completed. Whether a blocked step stayed blocked. Whether that record is independently verifiable after the fact.
Observability that only explains the incident after irreversible actions complete is not enough under this framing. The organizations that can produce write-time evidence of instruction, enforcement, and outcome are the ones positioned for the doctrine Ferguson just described out loud.
Source: Reuters