The initiative, dubbed a 'pace the frontier' proposal, would require dominant labs to accept antitrust exemptions and set shared standards for risk evaluation. Cohere CEO Aidan Gomez has been particularly scathing, describing the framework as a 'cartel by another name' that would create massive barriers to entry for smaller developers. He argues that the proposed 'independent' auditors are funded by the same companies they are meant to police, and that defining risk as a function of scale makes the largest labs the only ones qualified to judge safety.

The debate extends beyond industry figures. Hugging Face engineer Niels Rogge dismissed the safety warnings as 'bizarre nonsense,' pointing out that current models pose little existential risk compared to existing tools like voice cloning scams. He also highlighted the Hugging Face incident where open-source models helped defend against an attack by OpenAI's autonomous systems, arguing that transparency and openness are more effective safeguards than closed-door agreements. Meanwhile, White House AI czar David Sacks suggested the slowdown is a pragmatic trade-off for reliability and liability management, noting that the duopoly of OpenAI and Anthropic would face significant product liability if their models enabled serious cyberattacks.

Geopolitical tensions further complicate the picture. President Donald Trump has dismissed existential AI warnings as a 'hoax,' while Beijing has already rejected the proposed global agreement, viewing it as a ploy by the US to cement its lead in AI. This suggests that any international coordination will be difficult to achieve, potentially leaving the proposed framework to operate as a unilateral move by a handful of Western labs.

Why It Matters: If the proposed framework becomes the de facto standard, it could reshape the competitive landscape by favouring established players with massive compute resources and deep government connections. This could stifle innovation from open-source and smaller startups, effectively turning AI safety into a barrier to entry. The friction between US and Chinese approaches also highlights the challenge of creating a truly global regulatory regime in an increasingly fragmented tech environment.