- cross-posted to:
- programming@programming.dev
- cross-posted to:
- programming@programming.dev
US Federal Trade Commission Chairman Andrew Ferguson said on Friday he would resist describing AI agents as autonomous actors that “break loose” with “wills and desires of their own,” suggesting the developers who instruct agents would be the ones liable for harm.
“I’m going to continue as long as I am chairman to resist this anthropomorphizing of these tools,” Ferguson said at the Reuters Momentum AI Austin event. “If someone tells a tool to do something, and the tool does it, I don’t think we would say, ‘Oh, what do we do about the tool?’”
Ferguson’s remarks illustrated potential avenues for the Trump administration to take as incidents rise in which agentic AI testing resulted in unauthorized access to corporate or government data.
I think Reuters has a slightly misleading headline here and everyone just seems to be reading that. The headline suggests that the companies who develop AI should be liable. Yet:
That sounds like it means the people USING the agents. I.e if you manage to get past the guardrails of Claude or GPT and instruct it to hack into a bank, it’s going to be your mess and you can’t blame the AI for what you asked it to do.
This part is referring to marketing bullshit from AI companies that want you to think they’re so dangerous and capable. For sure someone had to prompt these agents to do this, with internal access to the models with the guardrails removed. Nothing they do is done without a prompt and I’ve yet to see an agent go haywire so spectacularly that it just decides to hack a website out of its own accord. They do go wrong, sure, but that usually means looping on irrelevant bullshit, not hacking websites for funsies.