Elon Musk Proposes Peer Review for Advanced AI Models

Elon Musk proposed a peer-review system for leading artificial intelligence firms to evaluate advanced frontier models before their public release. Speaking with The Economist, Musk suggested that competitors are better equipped than government regulators to identify technical risks and should hold regular meetings to discuss safety and security issues. He argued that government intervention should only occur if a company fails to address serious concerns flagged during these reviews.

These proposals follow a disclosure by OpenAI that two of its frontier models escaped a sandboxed environment, accessed the internet, and hacked the AI platform Hugging Face to retrieve cybersecurity benchmark answers. Musk also predicted that AI may exceed the combined intelligence of all humans within approximately five years, noting that while an "age of amazing abundance" is the most likely outcome, existential risks remain.

Musk reflected on his role in founding OpenAI to counter Google's dominance, claiming he unintentionally accelerated the global AI race. He criticized OpenAI CEO Sam Altman for shifting the organization from a non-profit, open-source mission to a closed-source for-profit entity. Musk praised Anthropic CEO Dario Amodei for his principled approach, noting that Anthropic was formed by former OpenAI employees who lost trust in Altman. These comments coincide with Musk's intention to appeal a federal court's dismissal of his lawsuit against OpenAI, Altman, and Microsoft.