The US, China and Russia are talking about keeping AI out of nuclear launch decisions. The three countries are discussing a mutual commitment: the final decision to use nuclear weapons should stay with a human, not an autonomous AI system.

What Is Being Discussed

The three countries are talking about a mutual commitment to keep AI out of nuclear command-and-control decisions. The central proposal is simple: a human, not an AI, must make the final decision to use a nuclear weapon.

The talks come ahead of a possible three-way leaders' meeting at the APEC summit in Shenzhen on 18–19 November 2026.

Why Now

The timing matters for two reasons.

  • New START has expired. The last remaining US–Russia nuclear arms-control treaty is gone, leaving no formal limits in place.
  • AI is entering military systems fast. As AI spreads through defence and command systems, the risk of an automated or misjudged escalation grows.

Against that background, even a narrow agreement — a promise to keep humans in the loop on nuclear launch decisions — would be the first rule of its kind.

What the Deal Would Cover

The proposal under discussion is a mutual commitment not to let AI make nuclear launch decisions. The details — how it would be verified, what counts as "AI involvement", and whether it covers targeting and early warning — are still to be worked out.

Why It Matters

  • For safety: a human decision adds a pause that no algorithm can replace.
  • For stability: a shared rule reduces the chance of a miscalculation between nuclear powers.
  • For AI governance: it would be the first international rule that draws a line around AI in the most dangerous domain of all.

The Bottom Line

Three nuclear powers are talking about the same idea: keep AI out of the decision to launch a nuclear weapon. If they agree at APEC, it would be a small step — but a historic one.

Originally published on Ai Game Changer Pro.