Fox Business Clips Sep 19, 2026
With Neil Chilson, former Chief Technologist at the FTC
Neil Chilson's answer to people who expect AI to slip human control is that a model's intelligence does not come with the power to act on the world.
The doomer case treats capability and control as one thing. Chilson pulled them apart: getting anything done outside the machine takes cooperation, and cooperation is on his list of what computers handle badly.
"These are powerful tools, don't mistake the ability for power, those are very different things"
Chilson, former Chief Technologist at the FTC, on Fox Business, held the technology post at the agency that brings cases against companies over how they handle data and software, which is where a rule of this kind would have to be enforced.
The full segment is covered here so you can skip it.
Here are the 3 takeaways that matter.
Key Takeaways
Intelligence is not power: acting in the world takes cooperation, which he says computers are not good at
An ordered slowdown would be impossible to enforce on other countries, and full of holes at home
The labs already watch their own models, so what they may need is more incentive, and he says liability can help
1. Ability Is Not Power
Asked what his message was to the doomers, Chilson did not dispute that the tools are powerful. He disputed the step from capability to control.
Being intelligent is not the same as being able to act
These are powerful tools, don't mistake the ability for power, those are very different things
Neil Chilson
Acting on the world takes cooperation and other work computers are bad at, he said, which makes safe development the job rather than containment of something already loose.
Build them carefully and the control problem does not arise
We need to develop them safely but we don't need to worry about them getting out of control if we do that well.
Neil Chilson
2. A Pause Nobody Enforces
The anchor asked whether one of the live regulatory ideas, ordering the companies to slow down or stop, could realistically be enforced. Chilson said it would mean standing behind every innovator and developer, and that other countries would keep building anyway.
Heavy-handed at home, unenforceable abroad
It would be extremely heavy-handed, that would be difficult to do, it would have a lot of holes in it. It would be impossible to enforce on other countries
Neil Chilson
3. Liability, Not a Stop Order
Told he read as a full-steam-ahead guy, Chilson pointed at what the companies already do. The precautions they take around testing, he said, are themselves an acknowledgment that they see risks, and he named OpenAI as a case of that acknowledgment. He compared the work to watching a pot on the boil instead of walking away from it for two hours.
Washington is not what makes the labs pay attention, but liability could add to it
The companies do not need the government to make them pay attention, Chilson said. What they may need is more incentive, and he said liability can help with that.
Bonus Insights
The anchor set it up as a machine with its own mind
He opened by saying AI seems to have a mind of its own, then asked Chilson for his message to the doomers who fear the threat it poses.
The sign-off went to the plaintiffs' bar
The anchor closed the segment by saying the liability lawyers will be all over it.
Chilson's bottom line is that the risk in AI is a question of how carefully it is built, not of losing control of it, and that an order to stop would bind US developers and nobody else.
Products, Companies & Tools Mentioned
OpenAI (Chilson named it as a company whose own steps acknowledge the risks, which is his argument for why the government is not the thing making the labs pay attention)
Listen to the full episode
🔴 YouTube
Watch the full episode:
More Neil Chilson interviews and podcast appearances
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: