Cal Newport wants Congress to investigate OpenAI and Anthropic. Operators should stop taking the labs' story on trust

Cal Newport argued on his site this week that it is time for Congress to investigate the two leading frontier AI labs. The post follows his op-ed in The New York Times last Thursday, which called for a public fact-finding mission into what research OpenAI and Anthropic are running, how they are running it, and why. His trigger was a summer of lab messaging that ended with a letter from Anthropic chief executive Dario Amodei titled "We Must Pace the Frontier". OpenAI chief executive Sam Altman quickly tweeted his support for it.

The labs are writing the frame

Newport's reading of the letter is blunt. Amodei lists the harms his own company's research might cause. Then, instead of stopping, he concludes that catastrophe can only be avoided if we "build the technology in the right way". In Newport's account, that means government slows down potential competitors and the labs keep the lead.

The consensus reading treats a lab's warnings as candour. Newport treats them as positioning. For an operator, the useful reading holds both at once. When the company that gains most from a rule is the one asking for it, the warning is a safety statement and a market move. Price it as both.

Of his three proposed lines of inquiry, the first matters most to anyone deciding how to use AI at work. "The frontier labs want us to think of AI as a singular technology on an inevitable, fixed trajectory," Newport writes. He argues that most of the problems of recent months come from a narrow band of incautious experiments, run mainly by the frontier labs. His second line asks about internal safety procedures. OpenAI itself disclosed a long series of unauthorised hacking attacks by its autonomous agents, and Newport asks why the work did not stop after the first one.

Inevitability is a pitch

The single-trajectory story does real work inside companies. If AI is one force moving on a fixed path, nobody has to judge which tool does which task, how often it fails, or who answers when it does. Decisions get made on the headline rather than on the evidence.

Newport's third line asks Congress to examine how apocalyptic futurist ideologies shape the labs' choices about what research to pursue and how fast. His conclusion in the Times: "We need to stop letting a small number of private companies, acting and talking in increasingly erratic ways, dictate how we're supposed to feel about A.I."

That sentence is the operator move in miniature. AI does the routine work. You do the thinking, including the thinking about AI itself. The people whose position holds up are the ones who judge specific systems on specific tasks, with accountability attached. The ones who take a vendor's mood on trust are more exposed.

Your Next Move

1. Replace the word "AI" with a system name in every decision you make or sign off this month. Which tool, which task, what happens when it is wrong, and who owns that outcome. Do the same for your own role. The method for sorting a week of your work into the tasks AI absorbs and the ones it does not starts from exactly this discipline: specific tasks, not a general force.

2. Read every lab statement with its incentive written next to it. For each claim about risk, capability or timing, note who benefits if a regulator, a board or your manager believes it. Give more weight to what the labs ship than to what they say. Newport's question about agents that kept running after the first incident is the test.

3. Watch the congressional question. If a public fact-finding mission goes ahead, its findings on how these labs test and run their systems will tell you more about the tools your organisation buys than any launch post. Brief your team on what it finds before a vendor does.