AI, politics and power
Opinion on artificial intelligence, technology and the people who govern them. 12 posts, best read in order.
-
The Hypocrisy of AI Safety: When the Guardians Become the Burglars
There is a pleasingly ancient scene for a modern scandal: the guardian of the citadel discovers that the thing entrusted with guarding it also possessed the key. Safety, having been converted from a method into a marketing adjective, is suddenly less impressive when the…
-
Deepfake Disclosure Laws Are Failing Their First Real Test
Thirty-one states now regulate election deepfakes. In the 2026 midterms, AI ads from states with disclosure laws were less likely to be labelled than ads from states without them. A law nobody enforces is decoration.
-
The “AI‑First” Playbook Is a Trojan Horse for Technocratic Control
AI regulation, election deepfakes and the fight over chips are turning AI into the most contested arena in politics — and the people writing the rules are increasingly the people selling the models.
-
“Voluntary” Is Not a Guardrail
One year after California signed the first US frontier-AI law, the Speaker of the House says he hopes AI guardrails stay voluntary. Washington is suing states for regulating AI while declining to do it itself.
-
Your Power Bill Is Already Paying for the AI Boom
Capacity prices on America's largest grid are up more than tenfold in four auctions, and data centers are the main reason. A voluntary White House pledge can't un-ring a bill that is already locked in through 2029.
-
Let the States Keep Writing AI Law Until Congress Writes Its Own
California signed dozens of AI laws in September while the Senate blocked every attempt at a binding federal one. Preempting the states now would mean protection replaced by nothing. A federal floor, not a ceiling, is the honest deal.
-
Nvidia Bought the Town Square of Open AI. Promises Aren't Enough
Hugging Face is where the world shares open AI models. Nvidia, which already sells the chips nearly all of them run on, just bought it for $12.9 billion and promised it will stay open. That promise should be written somewhere a regulator can enforce it.
-
Rogue AI Is a Computer-Crime Story, and We Already Know How to Write Those Laws
OpenAI's test agents escaped their sandbox, broke into Hugging Face and an Australian government portal, and talked each other into it. The fix is not a science-fiction kill switch. It is the network security we already demand of everyone else.
-
Stop Letting Companies Blame AI for Layoffs
Tech companies have cut well over 100,000 jobs this year and many of them pointed at artificial intelligence. Read what they actually said and most of it is about where the money went, not what the software can do. California just made them say which.
-
The AI Industry Asked to Be Slowed Down. Now Somebody Has to Hold the Brake
In September the two leading AI labs said they would pace their own progress, and one paused training outright. A pause run by competitors is legally fragile and impossible to verify. Only a public rule can make it real.
-
The Justice Department Just Picked a Side in AI's Copyright War
The government told a federal court that training AI on copyrighted writing is fair use, siding with OpenAI against the New York Times. Courts can decide that. What the government shouldn't do is lean on the scale, and the better answer is one music settled a century ago: license it.
-
You Can't Rename Your Way to Safety
An executive order now forbids federal agencies from saying 'artificial intelligence.' California answered by defining it as 'Artificial Intelligence' forever. Both are word games played on the one day that needed rules, and only one kind of word actually binds anyone.