OpenAI and Anthropic CEOs Urge UN to Establish International AI Safety Standards

OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei have called for stronger international AI safety standards during a United Nations Security Council meeting.

The two AI executives spoke separately at the meeting in New York on Wednesday. Although OpenAI and Anthropic compete in the AI industry, both leaders agreed that countries need common standards to manage the risks linked to rapidly advancing AI technology.

Altman urged governments to establish international standards for measuring AI capabilities, evaluating risks, and checking whether safety measures are working properly. He also called for meaningful human oversight as AI systems become more capable.

OpenAI and Anthropic CEOs Urge UN to Establish International AI Safety Standards

The OpenAI CEO said countries should quickly share accurate information about serious AI incidents. He argued that governments need to learn from failures before they develop into larger problems. Altman also proposed secure communication channels for governments to exchange information about emerging AI-related threats.

Altman stressed that major decisions about AI should involve democratic institutions and governments that are accountable to their citizens. His comments could create challenges for international cooperation because some of the countries developing advanced AI systems have different political systems.

Amodei presented several ideas for international cooperation. He suggested limited global agreements covering specific risks, including a possible ban on using AI to develop biological weapons.

The Anthropic CEO also proposed evaluation and verification systems that would allow countries to check whether others are following their commitments. He called for common standards for AI safety testing and an international system for reporting serious AI security incidents.

Amodei described AI as both a major opportunity and a significant global security challenge. He said countries should work together despite their differences to address the risks created by increasingly powerful AI systems.

Altman also warned that an AI-related catastrophe would not be acceptable. However, he said policymakers and the technology industry should avoid both extreme optimism and excessive fear about AI.

See Also: OpenAI Launches GPT-6 Sol and Luna with Better Performance and Lower Costs

He argued that focusing only on the possibility of an AI disaster could make it harder to address the technology’s potential benefits. AI could also help tackle major global challenges if it is developed and managed responsibly, he said.

The UN meeting is not expected to produce a formal international agreement. Instead, the discussions are intended to encourage countries to begin developing ideas for cooperation on AI safety, security and oversight.

The meeting highlights the growing debate over how governments should manage increasingly capable AI systems while allowing the technology to continue developing.