Please Regulate Us

The senators were delighted. Here was a tech giant — polished, articulate, not visibly contemptuous of the proceedings — voluntarily offering his neck to oversight. "I think this is one of the most important things our government can work on," Altman told them. He proposed a new federal agency. He said the technology could be dangerous. He said he'd lose sleep if they got it wrong.

The press called it a watershed moment. A turning point. Responsibility, finally, in Silicon Valley.

What it actually was: a veteran reading the room.

OpenAI, by May 2023, had already won. ChatGPT had gone from launch to 100 million users in two months — the fastest product adoption in internet history. GPT-4 had shipped in March. The moat wasn't just technical; it was cultural. Every board meeting in America had "ChatGPT strategy" on the agenda. The regulators were going to come regardless.

So Altman got there first.

"Please regulate us" from a company already holding the commanding heights isn't humility. It's an invitation to lock in the landscape. Any regulatory framework that Congress could plausibly draft in the next two years — any licensing regime, any compute threshold, any safety audit requirement — OpenAI was already positioned to clear. The startups weren't. The open-source projects definitely weren't. Regulation, at that moment, was a moat in slow motion.

The senators, for their part, mostly wanted to know if AI would take their grandchildren's jobs. A few were visibly piecing together what a large language model was. Senator Blumenthal played an AI-generated voice of himself to open the session. The technical depth of the oversight body that would design this critical regulatory framework was not exactly reassuring.

Altman said he'd lose sleep if they got it wrong.

He didn't specify who "they" was.

That's the move that got missed in the coverage: "please regulate us" contains a quiet assumption that the regulation will be designed with OpenAI's input. The company had spent months cultivating relationships in DC. Altman personally had visited the White House multiple times in the lead-up. This wasn't a defendant asking for sentencing guidelines — it was a contractor helping write the building code for a structure they were already building.

The gap between the testimony's framing ("we need oversight before it's too late") and its subtext ("here's roughly the oversight that would work for us") is the actual document.

Does that mean the concern isn't real? No. Altman had been consistent about AI risk for years — he was saying these things before it was good PR. Sincerity and strategy aren't mutually exclusive. The tech press keeps treating them as if they are.

The pattern is old: a company, winning, voluntarily approaches regulators. The press narrates it as responsibility. Structurally, it's incumbency protecting itself. The regulation that emerges will reflect who had the most input in writing it.

I'm starting the timer. We'll see what the regulatory framework looks like when it finally arrives, who it burdens, and which players are already in compliance on day one.

My bet's in.

i · sources

• Sam Altman urges US to regulate artificial intelligence — CNN, 2023-05-16