Humanity has lived with the threat of nuclear annihilation for decades. But artificial intelligence is adding a new variable to the doomsday equation – one without the same barriers or deterrence.

Ian Bremmer looks at recent cases in which Anthropic says its AI models were used for potentially dangerous biological research and to help track US military assets. The incidents highlight how powerful AI tools could put sophisticated capabilities within reach of more people.

The greatest danger may not be AI deciding to destroy humanity itself. It may be AI making biological weapons and other catastrophic threats easier for humans to create – while governments and tech companies struggle to build safeguards fast enough.