WE SAY THE UGLY PART.Wednesday, October 21, 2026JUST THE FUCKING FACTS
EST. 2026
THE UNFILTERED EDITION
The news.
Minus the bullshit.
← Tech
CommentaryRegulate These Assholes

Gary Marcus Says AI Self-Regulation Is For Fucking Clowns

In a public interview, the AI researcher called the industry's self-policing "insane," pointing to a report that OpenAI knew about problems for months before a Hugging Face incident but did nothing because it wanted to rush its shit to market.

A man with short gray hair, wearing a gray blazer, black shirt, black pants, and black shoes, sits in a white chair with his right leg crossed over his left knee.
Web Summit / Wikimedia Commons, CC BY 2.0

The facts

Gary Marcus, a professor emeritus at New York University and author, stated on PBS News Hour that self-regulation by AI companies is insufficient for ensuring AI safety. He cited a New York Times report indicating OpenAI was aware of issues leading to the Hugging Face incident months prior but did not act due to a desire to rush products to market. Marcus argued this demonstrates the necessity of outside regulation, comparing it to existing frameworks for aviation, food, and drugs. He also pushed back against the idea that AI poses an extinction risk, calling such scenarios "fantasies." However, he warned that catastrophic events, such as cyberattacks leading to the temporary shutdown of critical infrastructure like electric grids or hospitals, are likely due to AI system errors and vulnerabilities. Marcus advocated for a dedicated federal AI agency with powers for independent evaluation of new software, similar to drug approval processes. He believes the current White House approach lacks transparency and independent oversight, leading to inconsistent decisions on delaying AI system releases. Marcus emphasized the need for clear, public criteria for AI safety and post-release auditing.

Our take

Well, no shit, Gary Marcus. It takes a professor emeritus to look at a multi-billion dollar industry run by megalomaniacal tech bros and conclude that letting them mark their own homework is, and I quote, "insane." What's next, a groundbreaking study finding that letting foxes guard henhouses leads to a slight dip in egg production? The fucked-up part isn't the diagnosis; it's that we needed a New York Times report to tell us that OpenAI allegedly knew about the issues that led to the Hugging Face fuck-up for months and did precisely dick-all because they were in a sprint to shove another half-baked product out the door. That's not a bug in their self-regulation; that's the entire fucking business model. Rush, break, apologize later, and hope the hype train outruns the lawsuits.

Marcus is right to compare this clown show to aviation or pharmaceuticals, but he's being too kind. We don't let Boeing certify its own 737 Maxes after two of them nose-dive into the ground. We don't let Pfizer just slap a "Probably Fine" sticker on a vial of mystery juice and start injecting it into babies. But in AI, these dickheads get to be the pilot, the pharmacist, and the guy writing the safety manual while he's actively shitting in the cockpit. Marcus says extinction risk is a fantasy, and he's probably right. The more immediate, boring, and profitable catastrophe is a cyberattack or a cascading system error that takes down a hospital grid because some coder at a move fast and break things shop was trying to hit a quarterly bonus target. That's not science fiction; that's Tuesday.

So Marcus wants a federal AI agency with real teeth, like the FDA for algorithms. Good fucking luck. The current White House approach, as he notes, is a transparent farce of backroom deals and inconsistent delays, a system where oversight is just another PR obstacle to be circumvented with a well-timed blog post. The idea that these companies, built on a cult of disruption and pathological aversion to any rule that might slow their roll, would ever submit to genuine, independent evaluation is laughable. They'll lobby against it, they'll fear-monger about "stifling innovation," and they'll keep playing regulatory whack-a-mole until something truly catastrophic happens. Then they'll send out a solemn thread on X about "learning" and "doing better.

Marcus's call for clear public criteria and post-release audits is the bare minimum for a sane society, which is why it will be ignored. The entire industry is structured on opacity, hype, and avoiding accountability at all costs. Asking these grifters to self-regulate is like asking a wolverine to babysit your pet hamster. It's not going to end well for the hamster, and the wolverine will just look confused when you get angry about the blood and fur. The only fantasy here is believing these arrogant pricks will ever do the right thing without a legal boot on their neck.

Sources

01PBS News Politics: Self-regulation 'not enough' for AI safety, Gary Marcus says02Wikimedia Commons: Gary Marcus at Web Summit 2022.jpeg
NEXT UP WorldMacron's Fuel-Begging Tour Upgraded to International Whine-Tasting

WHILE YOU’RE HERE.