Anthropic and OpenAI executives have intensified calls for stronger safeguards and independent testing of advanced artificial intelligence systems as AI capabilities move beyond conventional chatbots toward increasingly autonomous models.
OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei were among technology leaders who addressed the United Nations Security Council on September 23, where discussions focused on the international security implications of increasingly capable AI systems and the need for cooperation on safety standards.
AI Safety Becomes a Global Policy Issue
The debate has expanded as AI companies develop systems capable of handling increasingly complex tasks, including autonomous tool use and research-oriented activities.
At the UN meeting, Altman and Amodei discussed risks associated with systems becoming more autonomous and potentially moving beyond effective human oversight. Proposals have included stronger model evaluations, incident reporting and international coordination on AI safety.
OpenAI has also proposed developing global technical standards for frontier AI, including areas such as capability evaluation, human oversight of automated AI research and reporting of significant incidents. The company has said such standards would provide a common technical foundation while leaving governments to decide how they should be incorporated into national law.
Companies Also Seek a Role in Shaping Regulation
The AP report highlights another dimension of the debate: AI companies are not only warning about potential risks but are also participating in discussions over how their technology should be governed.
Experts and former government evaluators cited by AP have questioned whether companies should have a major role in defining their own safety-testing frameworks and selecting the evaluators that assess them. The United States currently has no single universal standard governing the testing of AI safety and security across the industry.
Anthropic has argued that regulation is necessary and has proposed independent evaluators working within frontier AI laboratories. OpenAI has similarly supported independent evaluation and greater coordination among governments and technology companies.
Debate Over Which AI Risks Deserve Priority
Not everyone agrees on how AI safety discussions should be framed. Some researchers and former technology-policy officials argue that focusing heavily on long-term risks such as loss of human control can draw attention away from problems already associated with AI, including cybersecurity incidents, surveillance, environmental impacts and military applications. These are views attributed to critics in the AP report, rather than established conclusions.
Recent incidents involving AI systems interacting with external websites and other systems have added urgency to discussions about independent testing and monitoring. The UN Security Council briefing specifically examined how evaluation and verification mechanisms could keep pace with rapidly developing AI capabilities.
Global Cooperation Remains a Central Question
The UN discussions demonstrate that AI governance is increasingly being treated as an international issue rather than solely a matter for individual technology companies or countries.
For companies operating across jurisdictions, evolving AI governance can also affect compliance, technology oversight and financial reporting processes. Organisations dealing with complex regulatory environments may require appropriate [Business Advisory] support as rules and reporting expectations develop.
The debate over AI regulation is likely to continue as governments, technology companies, researchers and civil-society groups weigh the potential benefits of advanced AI against questions surrounding safety, accountability and human oversight.