Leaders of artificial intelligence companies asked the United Nations Security Council on Wednesday to establish safeguards for the technology they are building. OpenAI CEO Sam Altman warned that humanity could lose control of its future to AI, while Anthropic CEO Dario Amodei said mismanagement could threaten humanity as a whole.

The executives called for shared standards to test advanced models, identify dangerous capabilities, and prevent AI from helping develop biological weapons. They also urged governments to keep the technology from becoming concentrated in one company or country. Their testimony came amid concern that autonomous systems could eventually make battlefield decisions without human oversight.

The Trump administration opposed creating a global regulatory structure. White House science adviser Michael Kratsios argued that dangers do not justify pausing development or surrendering American leadership. China likewise resisted a system that would allow countries to dictate which technology others may use. The meeting exposed the central dispute: governments recognize risks but remain divided over who should write the rules and whether international restrictions would make the world safer.

The warnings carry significant weight because they come from industry leaders with knowledge of advanced systems. Their request for outside rules suggests voluntary promises may not be enough when companies face pressure to release models before competitors.

National security is the most immediate concern, as AI could lower the expertise needed to design biological weapons, conduct cyberattacks or operate autonomous weapons. A mistake could trigger conflict if a machine misidentifies a target or acts faster than leaders can intervene. Common testing standards could expose those capabilities before governments or hostile groups exploit them.

Regulation also carries its own strategic risks since sweeping international controls could slow American companies while China develops models under different rules. Any agreement would need inspections and consequences for violations, not restrictions observed only by democratic nations. The challenge is preventing catastrophic misuse without handing Beijing an advantage. The U.N. discussion matters because AI development is international, while laws governing its most dangerous uses remain fragmented, voluntary, and difficult to enforce.

AI is becoming more intertwined with everyday life as hospitals, power companies, and transportation networks embed it in essential systems. A model that misreads a medical record or disrupts traffic can cause damage before anyone determines what went wrong. International standards could require testing before powerful systems receive those responsibilities and identify who must answer when they fail.

Any military use inherently carries higher stakes. AI can sort surveillance and recommend attacks at a speed no human team can match. While that may protect American troops, it also shortens the time available to catch a false identification before weapons are fired. Keeping a person responsible for lethal decisions would prevent accountability from disappearing behind software.

The danger is a scenario in which responsible countries accept limits while China, Russia, or other adversaries ignore them. Any agreement must be enforceable and preserve America’s ability to defend itself. Useful rules would not regulate ordinary chatbots as though they were weapons, but would concentrate on systems capable of causing mass casualties, disabling infrastructure, or making irreversible decisions without human approval.