Jennifer Huddleston's position on the week's calls to slow AI down is that a company that wants to stop is free to stop — and that nothing about that gives a government the right to stop everyone else.
The argument over pacing the frontier has been conducted almost entirely in terms of risk. Huddleston's response is to ask what a pause costs on the other side of the ledger, and to point out that the harms people name are attached to particular uses rather than to the technology as a whole.
"When we're looking at government regulation, government regulation is a very static element."
Huddleston is a senior research fellow at the Cato Institute working on technology policy, and wrote the line the programme put to her at the top of the segment.
The full segment is covered here so you can skip it.
Here are the 6 arguments that matter.
👤 Guest: Jennifer Huddleston, Senior Research Fellow at the Cato Institute, who writes on technology and AI policy
🎙️ Host: Jonathan Ferro, who anchors Bloomberg Surveillance on Bloomberg Television
👥 Also on: Lori Calvasina of RBC Capital Markets and US Energy Secretary Chris Wright, in separate segments of the same programme
📰 Published: 14 September 2026 on the Bloomberg Surveillance podcast feed
🟣 Apple Podcasts | 🔗 Episode page | ⏱️ length not available
Key Takeaways
A company choosing to pause and a government mandating a pause are different things, and only one of them stops the market
Industry norms can adapt as risks emerge; a statute written today is frozen and can block the benign uses along with the risky ones
She calls the result a pacing problem
Most of the named harms — cybersecurity, fraud — are already illegal, and are attached to an application rather than to the technology
Electricity and water demand should start a conversation about the grid, not only about AI
By 2028 the harder question may be what still counts as AI, since nobody thinks of email autocomplete that way
1. Pausing Is Not a Mandate
Jonathan Ferro opened by reading Huddleston's own line back to her — that there is a significant difference between companies choosing to pause and a government mandate that could pause all AI development — and asked why, if they want to slow down, they do not simply slow down.
Her answer was that this is already an available option. "I think that's exactly one of the options."
She framed it as a right that belongs to the company. "If an individual company wants to not advance with their product, they have the right as an innovator, as an entrepreneur to say, this is where we're stopping."
What she expects to follow is competition, not a halt. This is a very competitive market, and others in it will carry on with the more advanced applications
The cost of a broad pause is the part she says is missing from the conversation. Most of the discussion has been about the risky impacts; against those she set the ability to respond to the same cybersecurity concerns, advances in fields like medical research, and the Department of Defense, which she said finds this research paramount to what it is doing right now
Ferro pressed the coordination problem: if every company wants an edge, does one holding back not require all of them to agree
2. Norms Move, Statutes Don't
Her answer to the coordination problem is industry norms. On specific issues such as cybersecurity, she expects norms and forms of self-regulation to emerge rather than top-down government regulation
The advantage she claims for that is speed. "One of the advantages of that, particularly in a fast-moving space like AI, is that it can be much more adaptive as new risks emerge, but also as new potential benign or beneficial elements emerge."
The contrast is the core of her argument. "When we're looking at government regulation, government regulation is a very static element." It takes a long time to evolve as the technology evolves, while AI is very dynamic
The failure mode has a name. "And so you could end up with a sort of pacing problem where even if technology is good and beneficial, it's unable to go forward if a government regulation is particularly outdated."
3. Congress or the States
On where any rules would come from, she pointed at Congress. "When we're talking about this type of major issue, you're likely to need some sort of congressional framework at some point."
What worries her more is the layer below it. "What's also been concerning is the number of states that have started to act around AI issues."
Asked about the risk of a state-by-state patchwork, she answered on the executive branch instead. Her point was that there have been various executive orders on AI and that they change significantly between administrations
She dated that pattern well before the current argument. "There have also been AI executive orders going all the way back to the Obama administrations."
4. Regulate the Application
Ferro put the spread of concerns to her directly — electricity costs, water usage, and now the eradication of humanity — and asked whether some AI advances should be treated like nuclear products or like a technology with growing pains.
Her first move was to narrow the object. Concerns about artificial intelligence are largely concerns about specific applications, and the harms people point to are tied to a particular use rather than to the technology as a whole
Her second was to note that most of it is already covered. There are existing laws in play around cybersecurity, around fraud, and around the range of cases where a bad actor could use the technology to do a bad thing
On electricity and water she wants the subject widened, not narrowed. She called those concerns a good opportunity to have a broader policy discussion about the underlying issues — the grid and water usage generally — rather than treating them as AI questions
5. Railroads, Not Warheads
Ferro sharpened the analogy: is this the growing pains of a new technology, like the proliferation of the railroads, or the kind of problem that produced nuclear treaties during the Cold War?
Her answer was that the analogy keeps failing because the technology is not one thing. "One of the things that makes artificial intelligence so unique is how general purpose of a technology it is." That, she said, is why several very different conversations run at once
She reframed it as an information technology. Artificial intelligence is an incredible tool for access to information, and there are many different forms of information one can access
Her closing question put the burden back on the regulator. What does it mean for the government to be the one that can dictate how this technology evolves, as against something that emerges from industry norms
6. What Counts as AI in 2028
Asked as a thought exercise what regulation looks like by the next presidential race, she started with the states. Whether a congressional framework and a single US approach to AI policy emerges is the first question
The second is what kind of approach it is. Either it continues the light touch she credits with letting the US excel in areas like the internet, or it is something much more regulatory that makes benign and beneficial applications harder to ship
Her most interesting answer was to the definition rather than the policy. As people get used to a technology, they stop calling it AI. "Very few of us think about the autocomplete on our emails as a form of artificial intelligence."
Her second example was the customer service chatbot that handles a request before a human picks it up, which nobody now classes as artificial intelligence either
Huddleston's bottom line is that the regulatory question is not whether to slow AI down but which specific application is being regulated, because a rule written about the technology as a whole will be outdated before it binds.
Bonus Insights
Ferro's own framing did work the guest did not have to do — he separated the risks into a quality-of-life category, a cybersecurity category and the comparison to a nuclear weapon, and asked her to say which kind of problem this is
The Department of Defense came up unprompted as a user of this research, which is an unusual thing to raise in an argument for lighter regulation and sets a national-security cost against a pause
She would not take the 2028 question as a forecast. Both times Ferro asked her to look ahead, she answered with what will be interesting to watch rather than with a prediction
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: