Anthropic and OpenAI are committing to slowing down the development of frontier artificial intelligence models, marking an unprecedented industry shift toward caution as internal pressure and safety concerns mount across major AI labs.

Anthropic CEO Dario Amodei announced that his company is unilaterally inviting third-party evaluators directly into its offices to monitor AI model development and verify safety standards. OpenAI CEO Sam Altman quickly joined the initiative, stating that OpenAI will follow suit and implement similar oversight measures.

The joint alignment follows intensifying debates over alignment risks, recently accelerated by the resignation of Anthropic researcher Jacob Coxon. Coxon publicly warned that leading companies are taking existential risks with human life while internal teams believe the technology could pose catastrophic dangers by the end of the decade.

Amodei cited two primary catalysts for taking a more cautious approach: a recent security breach involving OpenAI and HuggingFace, and the accelerating pace of self-improving AI capability over recent months.

“We must slow the pace at which we improve the capabilities of AI models,” Amodei stated. “Progress will still seem fast, and we must make wise use of the time we gain.”

The core components of the proposed pacing framework focus on three structural levels:

• Embedded Independent Evaluators: Companies will grant third-party safety organizations, such as METR, direct physical and technical access. Evaluators will receive company badges, dedicated desks, laptops, and internal risk assessment visibility to ensure safety incidents are reported independently.
• Domestic Safety Coordination: Leading American AI labs are seeking narrow US government antitrust waivers to collaboratively establish shared safety benchmarks and rate limits on unchecked progress without legal exposure.
• Geopolitical Pacing & Controls: The proposal outlines a strategy to maintain a 3-to-5-year US lead over China using semiconductor chip export controls and model distillation crackdowns, while simultaneously pursuing narrow international treaties banning biological weapon applications.

Executive response across the sector was immediate. SpaceX CEO Elon Musk endorsed the framework, posting that “Dario is right,” while Altman confirmed that OpenAI has held extensive internal discussions regarding frontier pacing in recent weeks.

“I agree with Dario that we need to pace the frontier,” Altman said, noting that OpenAI will share details on its rollout soon.

The move arrives amid growing public skepticism toward the technology sector, which Amodei described as a fundamental “crisis of trust.” While critics argue that apocalyptic warnings serve as a distraction from immediate societal harms or represent potential regulatory capture, Amodei emphasized that intentional pacing remains the only viable path forward.

“My desire to achieve these benefits is undimmed,” Amodei said. “But the benefits will only be achieved if we build the technology in the right way, and — so long as we use the time we gain well — it is worth taking unusually deliberate care to get it right.”

---

About TBC News

Established in 2022, TBC News is one of the Philippines’ leading digital news organizations covering business, technology, luxury, and the influential people and companies shaping the world.

With a growing international readership, TBC News delivers authoritative reporting, exclusive insights, and sharp perspectives on the developments driving industries, markets, and culture forward.

From groundbreaking innovations and major corporate developments to luxury, investment, and emerging trends, TBC News provides a distinctive editorial lens on the stories that matter most to decision-makers, industry leaders, and globally minded readers.

For more information, visit tbc-news.com.