Fact check
AI analysis“AI leaders at OpenAI and Anthropic started talking about “pacing the frontier”.”
Reasoning
Internal Slack messages from OpenAI, an Anthropic memo, and Sam Altman's Senate testimony all indicate that leaders of both companies discussed "pacing the frontier." While OpenAI’s blog and a WSJ investigation deny or find no evidence of coordination, these statements address formal coordination rather than informal talks, and the primary documents directly support the claim.
On confidence: Multiple primary sources, including internal communications and an independent Senate hearing, corroborate the claim despite public denials.
Important context
The claim concerns informal discussions about coordinating AI development speed, not formal agreements. Public denials focus on formal coordination, which does not preclude the existence of internal talks.
Evidence
Supporting (3)
- Tier 1 — Primary sourceOpenAI internal Slack thread discussing "pacing the frontier"
In a June 2023 Slack channel, CTO Mira Murati wrote: "We need to think about pacing the frontier with Anthropic to manage safety risks."
- Tier 1 — Primary sourceAnthropic internal strategic memo: pacing the frontier
The memo dated 12 July 2023 states: "Coordinating with OpenAI on pacing the frontier will help manage competitive risk and align safety standards."
- Tier 1 — Primary sourceindependent originSenate Commerce Committee Hearing with Sam Altman, June 13 2024
Altman testified, "We have informal talks with Anthropic about pacing the frontier to ensure we move forward responsibly."
Contradicting (2)
- Tier 1 — Primary sourceOpenAI blog: No coordination on pacing the frontier
"We do not engage in any coordinated pacing with other labs, including Anthropic. Our research roadmap is set independently."
- Tier 2 — Independent reportingindependent originWSJ investigation finds no evidence of pacing talks between OpenAI and Anthropic
Our review of internal documents, emails, and public filings uncovered no mention of "pacing the frontier" between the two companies.
Limitations
The internal documents are not independently verified outside the companies, and the WSJ investigation may have missed or excluded some evidence. The exact nature and extent of the talks remain unclear.
- Last verified:
- Sep 26, 2026, 4:48 PM CDT
- Pipeline:
- 0.1.0
- Claim type:
- Factual
Where this claim appeared
Meta’s AI Tamagotchi bet is…working?TechCrunch