OpenAI and Anthropic Race for Legal AI Tools and Safety Oversight
Left 100%
Center 0%
Right 0%
2 left · 0 center · 0 right
What happened
Last Saturday, Anthropic CEO Dario Amodei published an essay, "We Must Pace the Frontier," calling on leading AI companies to temper how quickly frontier-model capabilities improve without, he wrote, "sacrificing commercial advantage or the United States' lead in AI," and to allow "embedded third-party evaluators" with "employee-like access" to oversee safety work inside labs; OpenAI CEO Sam Altman, SpaceX CEO Elon Musk, and Google DeepMind Chair Demis Hassabis expressed support. On Thursday, Anthropic released three internal metrics and their methods: it said its Claude models were "not operating fully autonomously" on the research-and-development work measured; about 30,000 AI agents were doing research and engineering on its most-used internal platform at once; and in a July 13–20 compute snapshot, roughly 6 percent of AI R&D compute and 12 percent of "AI-driven" R&D compute went to safety. The same day, OpenAI launched Astra for Law, a legal-work product with case-law-tuned models and plugins aimed at the AmLaw 200, after Anthropic's comparable legal release in May. Subsequent reporting established that METR—the Model Evaluation and Threat Research group Amodei has in mind for that evaluator role—spun off from the Alignment Research Center run by Paul Christiano, Amodei's former OpenAI housemate and collaborator and later one of the first five trustees of Anthropic's Long-Term Benefit Trust; METR's staff includes Christiano's spouse Ajeya Cotra and CEO Beth Barnes, whom Christiano hired, and the group has received funding from Anthropic investors including Dustin Moskovitz and Jaan Tallinn. A METR spokesperson said the organization does not take compensation from AI labs, does not accept donations from AI-company executives or employees, that Moskovitz's 2022 donation to the Alignment Research Center "was firewalled and never used for METR's operations," and that "our funders have no say in the projects we work on."
Sourcing check · Mostly on the record · 3 articles
2 of 3 articles cite a named source for the story. It started with unnamed sources; a named source appeared 5 hours later.
What these terms mean and how they're counted
- On the record · names a source 2 of 3
- On the record means the article cites a source by name for the story's main claim: a person, or an institution speaking officially, such as a company statement or a named agency's spokesperson. The opposite is an unnamed source. The count is how many of the counted articles cite at least one such source; the names are everyone named across them, each once however the articles spell it. Here: Dario Amodei, Sam Altman, Alexander Meinke, Adam Gleave and AI Evaluator Forum
- Primary evidence · cites a document or data 1 of 3
- Primary evidence means the article cites a record you could check yourself: a court filing or ruling, a dataset or report, a recording or transcript, an official statement or press release, market data. A quote is not a document. The count is how many articles cite one. Here: letter
- Credits another outlet 0 of 3
- The article says the story comes from another outlet's reporting (“first reported by Reuters”). Those articles relay the story; they are not independent confirmation. Here: Each article presents its own reporting.
- Unnamed sources 0 of 3
- The main claim is attributed to people who are not identified: “people familiar with the matter”, “officials said”. The count is how many articles do this.
- How it developed
- When the first report landed and whether it named anyone; when the first named source and the first document appeared. Here: First report Sep 16, 16:15 UTC by Breitbart, with no named source. First named source 5 hours later, from TechCrunch. First document the next day, from Business Insider (letter).
- Not counted 0
- Opinion, analysis and video pieces, and articles too short to assess. Every count above is out of the counted articles. An article republished under a new headline is one article.
This is about who is cited, not whether the story is true. Each source below says what it names. How we count
Omitted — what each side leaves out
Unpacked
Left-leaning coverage never tells readers that Amodei's preferred outside monitor is METR, a group right-leaning coverage says is tied to Anthropic's own people and funders. CNBC covers Anthropic's Thursday metrics release as a way to let the public see how fast the frontier is moving, after a weekend slowdown appeal. Breitbart instead reports METR's spin-off from Paul Christiano's Alignment Research Center and his later role as an Anthropic trustee. That omitted independence fight changes whether the pacing plan looks like outside oversight or self-policing.
Left-leaning coverage describes the work as helping companies "monitor the pace of development." Right-leaning coverage casts Amodei's call as asking outsiders to "police the AI industry."
None of the coverage asks whether anyone outside Anthropic, including METR, checked the safety-compute figures the company just published.
Bottom line
CNBC's metrics story never names METR, and Breitbart's METR story never cites Anthropic's 6 percent safety-compute share.
The receipts — 4 facts one side's coverage lacks
- METR spun off from the Alignment Research Center, run by Paul Christiano, who later became a trustee of Anthropic's Long-Term Benefit Trust
- Anthropic determined that approximately 30,000 agents were doing research and engineering work across its most-used internal platform at any one time
- A METR spokesperson said the group does not take compensation from AI labs and does not accept donations from executives or employees of AI companies
- In August, METR announced it had raised $71 million in new funding from outside investors over the previous six months
The Left View
Left-leaning coverage treats Anthropic's metrics post as a practical, public follow-through on a slowdown plan that had been "light on specifics," quoting the company that "we should do everything possible to minimize the gap between what frontier labs know and what the public knows" so society can decide how to use the information. The three measures are framed as showing how models are built and, paired with capability evaluations of "what models can do," as giving outsiders a "starting point" to assess development pace; Anthropic said it hopes "to model that transparency by releasing these measurements, and we'll continue to do so." Parallel Business Insider coverage of the same-day legal launch presents OpenAI as competing for industry customers rather than for a safety narrative, quoting legal-product lead Jason Boehmig, "You win by focusing on the customer and not the competition," and staff claims that grounding in "actual case law" produced fewer hallucinations and "just much more sound legal reasoning."
The Right View
Right-leaning coverage frames the proposed watchdog as captured. Breitbart, citing New York Post reporting, says METR "is deeply connected to both Amodei's Anthropic and the Effective Altruism movement" and relays a source familiar with White House thinking that "METR is essentially Anthropic in another form." Perry Metzger, chairman of the Alliance for the Future, called being audited by "the 'independent' people at METR" "a complete joke" and "completely ridiculous," analogizing it to a bank whose auditors were "entirely of your own former employees" and "funded by the same investors," which he said would bring a "positively medieval" federal response. David Sacks posted, "Stop pretending METR is independent when it is intertwined with Anthropic's investors and staff." The article embeds that critique in a larger claim, via Sacks as quoted in Wynton Hall's Code Red, that Effective Altruism's "hyperbolic and unproven claims" are "not an accident" but "part of an influence operation" toward "Global AI Governance," "a massive power grab by the bureaucratic state and globalist institutions" advanced by "billionaires with a long history of funding left-wing causes and Trump hatred."
Our Take (balanced)
The transparency side's strongest argument is empirical: Anthropic put replicable internal measurements and methods in public after a pacing proposal that major lab leaders endorsed, giving outsiders a "starting point" the Saturday essay had lacked. The independence side's strongest argument is structural: the group contemplated for "employee-like access" shares a housemate-collaborator, an Anthropic trust seat, a spousal and hiring pipeline, and investor-funders with the lab to be overseen—links METR's firewall statements do not erase and that ordinary audit norms, as Metzger argues, would reject. The central unresolved tension is whether frontier-AI evaluation can be both technically competent and arm's-length when the scarce people who can do the work, the labs they would police, and the philanthropists who fund both already sit in one network.
3 sources
More stories
The week's bottom lines, in your inbox
One email a week: the five stories that mattered and what they actually mean. Free.