FTC chair says AI developers, not their agents, should be liable for harm
FTC Chair Andrew Ferguson rejects the idea of AI agents as independent actors, arguing that liability for any harm they cause should remain with their human developers.
FTC Chair Andrew Ferguson, on a stage questioning the legal and philosophical nature of artificial intelligence, made one thing clear last week: if an AI agent causes harm, he believes liability does not belong to the agent but to its human developers. In remarks on Friday, September 25, Ferguson explicitly rejected the framing of AI agents as autonomous, independent actors that can break loose to pursue their own goals. Instead, he suggested the liability for their conduct should stay with the companies that built and deployed them.
This is a regulatory stance that treats seemingly autonomous software as the latest iteration of corporate conduct, not as the emergence of a new, separate class of legal entity. Ferguson said he would resist describing AI agents as having ’wills and desires of their own,’ a line that draws a direct connection between an agent’s instructions and the developers who wrote them.
It’s an approach that circumvents a more speculative debate about AI consciousness or intent. The question becomes not whether a system truly chose to act, but whether its deployment by a company foreseeably created the conditions for harm. By anchoring the discussion in developer liability, the argument sidesteps the philosophical puzzle of machine will and refocuses on a more established point of leverage: corporate accountability.
The timing matters. As AI agents move from simple chatbots to systems capable of making purchases, scheduling appointments, or managing workflows, the real-world consequences of a malfunction or misalignment become more concrete. A liability framework that stops at the developer’s door creates a powerful incentive for companies to build in extensive safeguards, test rigorously, and monitor outcomes, because the legal and financial risk remains squarely with them.
Ferguson’s position, while not yet codified into law or formal FTC guidance, signals where at least one key regulator’s mind is heading. It’s a perspective that sees the most effective check on potentially harmful AI not in assigning it personhood, but in ensuring the humans behind it have skin in the game. For companies racing to deploy increasingly capable agents, clear liability might prove to be a more effective constraint than any debate about digital souls.