Who will decide that level AI you have access to?

Anthropic demonstrated that its Claude AI models can accelerate drug discovery by designing binder molecules and performing chemical analysis more efficiently than manual methods. However, the company plans to restrict access to its most advanced models to select scientists. Meanwhile, OpenAI's CEO Sam Altman announced a pause in frontier reinforcement learning training to focus on alignment and security, suggesting a cautious approach to releasing powerful AI capabilities.