OpenAI CEO Sam Altman has publicly agreed with Anthropic’s Dario Amodei on the need to slow down the pace of AI development while increasing transparency through independent safety assessments. This alignment marks a significant step in the industry's approach to managing AI risks.
- OpenAI will grant independent evaluators access akin to employees for safety checks.
- Anthropic and OpenAI endorse slowing AI development to enhance safety measures.
- Europe lacks mechanisms for competitor safety collaboration due to past policy reforms.
What happened
Sam Altman, CEO of OpenAI, has endorsed a proposal by Dario Amodei of Anthropic calling for the AI industry to slow its rapid development pace. This agreement follows Amodei’s advocacy for allowing independent evaluators employee-like access within AI labs to verify safety measures firsthand. Both leading AI companies have now publicly committed to this enhanced transparency approach.
Amodei also urged U.S. regulators to consider a narrow antitrust exemption that would allow AI competitors to hold discussions focused on safety without legal repercussions. However, this request highlights regional differences, as the European Union abolished such exemptions in 2004 and currently does not offer a similar framework for safety collaboration among competitors.
Why it matters
The alignment between Anthropic and OpenAI on slowing down AI progress signals recognition among top labs that unchecked rapid advancement could pose significant risks. Providing independent evaluators with employee-like access offers a novel approach to transparency and accountability in AI safety without compromising intellectual property.
The proposed U.S. antitrust waiver would enable cross-company safety coordination, filling a regulatory gap that currently hampers cooperative safety actions. Europe's lack of such a waiver underlines differing regional approaches, with Brussels emphasizing past reforms that restrict individual exemptions, although it has developed soft safe harbors in other areas like sustainability standards.
What to watch next
Industry and regulators will be monitoring how OpenAI implements its commitment to external evaluation access, including whether other AI developers follow suit. The practical impact of this approach on enhancing AI safety and public trust remains to be seen, alongside how these evaluator teams operate within competitive labs.
U.S. authorities will need to consider Amodei’s call for antitrust waivers that facilitate collaboration on AI safety. At the same time, Europe may revisit its horizontal cooperation guidelines or push for global frameworks addressing AI safety standards, recognizing the challenges posed by the absence of exemptions for competitor discussions on this front.