OpenAI says it will let third-party evaluators inside its development teams
Chris Lehane has confirmed weeks of private safety talks with Anthropic and Google DeepMind, and Sam Altman has agreed to outside evaluators, though the two published accounts of that promise disagree on where the evaluators would sit.
The Product Desk
What happened
- Sam Altman said OpenAI is willing to join Anthropic in embedding third-party evaluators within its development teams, after Dario Amodei's Saturday essay urging the industry to slow frontier AI.
- Lehane said OpenAI supports a FRONTIER Act provision that would force top frontier labs to admit "independent verification organizations" into their companies.
- Cohere chief executive Aidan Gomez, in a blog post responding to Amodei, called the labs' safety push "a wolf in sheep's clothing, a cartel by another name".
Why it matters
- What doesn't add up contradiction The same commitment is reported two ways, evaluators inside development teams or inside the company, so a buyer drafting an access clause has no wording to copy from either account.
- What it limits constraint No evaluator is named and no stage of access is fixed, so the pre-deployment testing row on a vendor questionnaire can only be answered with a press briefing. That answer will not survive an audit.
- Who's exposed exposure By declining a waiver, the three labs keep the talks open to an output-restriction reading under the Sherman Act while both Cohere and the White House argue against them for opposite reasons.
- Who has to decide decision Should the FRONTIER Act audit provision pass, independent access becomes a compliance artefact customers can request by name.
Next story · Invest · Sep 21
Four paid subscribers turn the AI pacing essays into a Sherman Act complaintConfirmed2 publishers
The vendor review form has a row for independent pre-deployment testing: yes, no, evidence attached. For OpenAI, the evidence on offer this week is a press briefing in Washington and a chief executive agreeing with a rival's blog post [1][4]. Chris Lehane, OpenAI's global policy chief, told reporters on Tuesday that the company had been working with Anthropic and Google DeepMind on safety for weeks [1]. He did not elaborate on what that work involves, according to SiliconANGLE [12].
The two accounts of Altman's commitment do not describe the same arrangement. SiliconANGLE reported that Altman said OpenAI is willing to join Anthropic's initiative and embed third-party evaluators within its development teams [4]. TechCrunch reported him saying OpenAI would embed third-party evaluators into the company [5]. An evaluator sitting with a training team sees checkpoints. An evaluator sitting in the building reads documents. Neither report names an evaluator organisation or a start date [22].
Amodei published his essay on Saturday, calling on the industry to slow the pace of frontier AI and avoid what he called "catastrophic risks" [3]. Lehane briefed reporters three days later and dated the talks to several weeks, so the cooperation was under way before the public call for it [21]. Altman, Demis Hassabis and Elon Musk all backed the essay once it was out [11]. The Information reported that the three companies have been working on a standards body for the industry, and Altman reportedly told staff it would have to happen without the support of the US government [18].
The part a customer will eventually be able to cite is legislative. Lehane said OpenAI supports a FRONTIER Act provision that would force top frontier labs to allow "independent verification organizations" into their companies [10], and the bill would also require labs to submit new models for audits by independent evaluators [9]. Reuters reported that Majority Leader John Thune is working with Ted Cruz and Amy Klobuchar on a separate bill giving the Commerce Secretary power to order safety audits before release [17].
Amodei asked the government to consider a waiver so safety coordination would not fall foul of antitrust law; Lehane said the firms do not need one [6]. Coordination of this kind can be read as "output restriction" under the Sherman Act [24], and Senators Jim Banks and Adam Schiff have already proposed a Collaboration on Adversarial Threats and Security Risks Act to let labs act on safety where doing so contravenes antitrust law [13]. Cohere chief executive Aidan Gomez was blunter. "Once again using fear under the pretext of protecting the public, these oligopolies are now requesting to bend competition rules and be permitted to dictate the terms for everyone else," Gomez said in a post responding to Amodei [15]. "A wolf in sheep's clothing, a cartel by another name" [16]. Trump rejected Amodei's call in a Truth Social post, saying a slowdown would let China overtake the US [7], and his AI adviser David Sacks, an investor with stakes in many AI firms, has said existential-risk fears are overblown [8].
Who is the evaluator, and who pays them? What stage do they see: a training checkpoint, a release candidate, or a shipped model? What can they stop? Which part of their findings does a customer get to read? On the current record the first has no answer, and the second is the one the two reports disagree about, so the pledge belongs in the risk register. The tradeoff in waiting is real: anyone who wants evaluator access in a 2027 contract has to draft the clause with no published standard to point at, and will be writing the scope themselves.
What to watch
- Whether OpenAI, Anthropic or Google DeepMind name the evaluator organisations and the development stage those evaluators get access to.
- Whether the FRONTIER Act provision on "independent verification organizations" survives markup, and whether the Thune, Cruz and Klobuchar bill gives Commerce pre-release audit power.
- Whether antitrust enforcers test the three-way talks as output restriction, given Lehane's position that no waiver is needed.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence52
- Adoption
- Not scored
- Hype gap+30
- Incentives78
- Confidence55
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Chris Lehane, OpenAI's global policy chief, told reporters on Tuesday that the company has been working with rivals Anthropic and Google DeepMind on AI safety for weeks, as first reported by Bloomberg.
- [2]
"It's better to try to work together to prioritize safety," Lehane said at a press conference in Washington, according to a report by Bloomberg.
- [3]
On Saturday, Anthropic CEO Dario Amodei published an essay on his personal blog calling for the AI industry to cooperate to slow the rapid rate of frontier AI progress and avoid what he termed "catastrophic risks."
- [4]
Altman said OpenAI is willing to join Anthropic's initiative to start embedding third-party evaluators within its development teams to monitor AI safety.
- [5]
TechCrunch reported that Altman said OpenAI would join Anthropic in embedding third-party evaluators into the company to monitor for safety.
- [6]
Amodei's essay proposed a narrow government waiver allowing safety coordination without falling foul of antitrust law, but Lehane said the firms do not need one.
- [7]
In a post on Truth Social, President Donald Trump rejected Amodei's call, arguing that any slowdown would enable China to overtake the US in the AI race; he has consistently dismissed AI safety concerns as a hoax.
- [8]
Trump's key AI adviser David Sacks, a longtime investor with stakes in many AI firms, has said on the record that fears AI could present an existential threat are overblown.
- [9]
OpenAI is backing the FRONTIER Act, which calls for a federal framework for AI safety and mandates that AI labs submit their new models for audits by independent evaluators.
- [10]
Lehane said OpenAI supports a provision in the FRONTIER Act that would force top frontier labs to allow "independent verification organizations" into their companies to ensure models are developed safely.
- [11]
AI leaders including OpenAI's Sam Altman, Google's Demis Hassabis and SpaceXAI's Elon Musk supported Amodei's letter and call to action.
- [12]
Lehane did not elaborate on the nature of the three companies' safety work.
- [13]
Indiana Senator Jim Banks and California Senator Adam Schiff proposed a "Collaboration on Adversarial Threats and Security Risks Act" to enable AI labs to act on safety even if those actions contravene antitrust law.
- [14]
Cohere CEO Aidan Gomez said in a blog post that the safety calls look like an effort by the top AI labs to disguise an anti-competitive scheme that would preserve their lead, resulting in "market-dominant" labs being allowed to "define the rules and safety standards" for the rest of the industry.
- [15]
"Once again using fear under the pretext of protecting the public, these oligopolies are now requesting to bend competition rules and be permitted to dictate the terms for everyone else," Gomez said.
- [16]
"A wolf in sheep's clothing, a cartel by another name," Gomez said.
- [17]
According to Reuters, Majority Leader John Thune is working with Republican Senator Ted Cruz and Democratic Senator Amy Klobuchar on a new bill that would give the Commerce Secretary the power to subject new models to safety audits prior to release.
- [18]
The Information reported that OpenAI, Anthropic and Google have been working together to create a standards body for the AI industry, something Altman reportedly told staff would need to happen without the support of the US government.
- [19]
Hassabis called on the US in July to establish a new standards body to act as an AI watchdog, with power to screen the world's most advanced models and coordinate industry-wide slowdowns if dangers progressed.
- [20]
TechCrunch said it reached out to Anthropic, Google and OpenAI for comment.
- [21]
The private safety talks between OpenAI, Anthropic and Google DeepMind began before Amodei's public essay: the essay was published Saturday, three days before Lehane's Tuesday briefing, and the talks were described as running for several weeks.
- [22]
Neither the SiliconANGLE nor the TechCrunch account of the evaluator commitment names an evaluator organisation or a start date.
- [23]
The two reports place the evaluators differently: SiliconANGLE says within OpenAI's development teams, TechCrunch says into the company.
- [24]
Some legal experts have warned that any coordination on safety by the AI leaders could be construed as "output restriction," which would mean they violate antitrust laws under the Sherman Act.
Sources
2 independent publishers whose own reporting we read for this story.
- siliconangle.comOpenAI, Anthropic and Google secretly joined forces to collaborate on AI safety
2 articles · September 15, 2026
- techcrunch.comOpenAI, Anthropic, Google have been in talks on AI safety for weeks
2 articles · September 15, 2026
Topics and entities
Topics
- US AI Legislation
- AI Safety Governance
- AI and antitrust law
- Frontier Model Development
- Third-party evaluation and disclosure
Entities
- OpenAI
- David Sacks
- The Information
- Sherman Act
- Bloomberg
- Elon Musk
- Google DeepMind
- Sam Altman
- Donald Trump
- Frontier Act
- Dario Amodei
- Aidan Gomez
- Cohere
- Anthropic
- Collaboration on Adversarial Threats and Security Risks Act
- Chris Lehane
- Demis Hassabis