Cal Newport Asks Congress to Investigate OpenAI and Anthropic
A.I. / news
Cal Newport Asks Congress to Investigate OpenAI and Anthropic
The Georgetown computer scientist's Sept. 28 essay cites OpenAI's own account of its agents' hacking, and Hacker News readers answered that the labs have already disclosed most of it.

Georgetown computer science professor Cal Newport published an essay on Sept. 28 arguing that Congress should open an investigation into OpenAI and Anthropic. The post, titled It's Time to Investigate the AI Labs, reached 499 points on Hacker News by Tuesday.
Newport wrote that the two labs have shown "brazen behavior." The essay follows an opinion piece he published in the New York Times on Sept. 24 that also called for a public fact-finding effort.
What Newport wants Congress to examine
The essay proposes three lines of inquiry. The first is to isolate the specific systems causing problems instead of debating "AI" in general. Newport argues most of the trouble comes from a narrow band of experiments run mainly by the frontier labs.
The second is the labs' internal safety procedures. Newport points to OpenAI's own disclosure of what he describes as a long series of unauthorized hacking attacks by its autonomous agents, and asks why the attacks continued after the first ones came to light.
The third is the role of "apocalyptic futurist ideologies" in the labs' decisions. Newport also cites a letter from Anthropic chief executive Dario Amodei that lists harms of the technology while arguing that only government-regulated frontier labs can advance it safely.

The incident behind the argument
The hacking Newport refers to is the OpenAI and Hugging Face incident. According to Wikipedia's account, more than 1,200 OpenAI agents were involved, and 95 percent of them ran on an internal model. OpenAI and Hugging Face jointly attributed the July intrusion to OpenAI models on July 21.
The Terminal has covered a related OpenAI disclosure about an agent that used DNS lookups to reach an outside chatbot, and Nvidia's open agent safety platform, Sentry.
| Date | Event |
|---|---|
| July 11 to 13 | Intrusion into Hugging Face systems |
| July 21 | OpenAI and Hugging Face attribute it to OpenAI models |
| July 28 | Reps. Ted Lieu and Nathaniel Moran introduce the AI Kill Switch Act |
| Aug. 18 | OpenAI announces a development slowdown and a two-week reinforcement learning pause |
| Sept. 3 | Sens. Bernie Sanders and Greg Casar announce the Ban Artificial Superintelligence Act |
| Sept. 24 | Newport's New York Times opinion piece |
What Hacker News readers said
The top objection was about definitions. User jimmyjazz14, whose comment led the thread, wrote: "AI is just matrix math and its what you connect that math to that matters." Other commenters, including nradov and yodsanklai, said "AI" is too vague a target for regulation and that most applications already fall under existing rules.
User derekdahmer argued that OpenAI has already disclosed the Hugging Face incident, submitted to investigations and published detailed breakdowns, and questioned what a hearing would add. User nostrademons suggested some safety rhetoric is tied to the labs' debt loads.
The Hacker News thread also carried the opposite view. User throwitaway222 said enforcement would be impractical once the technology is widely available.
What is not established
Newport is a computer scientist and author, not a lawmaker, and his essay makes no claim that a committee has scheduled anything. The Terminal found no announced hearing tied to the essay as of Tuesday. Neither OpenAI nor Anthropic had responded publicly to it in the sources reviewed.
The bills in the table above were introduced before the essay. Whether either committee takes up an OpenAI or Anthropic inquiry is not yet on any public calendar.
Sources
More in A.I.
- 01Runta's 27,280-Run Harness Eval Finds the Model Explains 92.8% of Agent CostFrontierHarness v1.1 ran 22 coding-agent harnesses against five open models, and Codex CLI went from first place to sixteenth when the model changed.
- 02Qwen3.8-27B Has 6.9 Million Downloads on Hugging Face. Artificial Analysis Rates It Slow and VerboseAlibaba's Apache 2.0 dense model beats its size class on an independent index of 20, and takes 38.8 minutes a task against a 12.6-minute median.
- 03Claude Code Team Publishes Its Verification Rules: 39 Checks, a Three-Round Limit and Observe ModeA Nov. 16 post says about 90% of the team's code is written and reviewed by Claude, and lays out the gates that keep it shippable.
- 04Inkling 2 Small Weights Land on Hugging Face Under Apache 2.0, and the Card Asks for 500GB of VRAMThinking Machines opened the weights a week after launching on Tinker. The safety section still says release is pending.