AI is becoming introspective – and that ‘should be monitored carefully,’ warns Anthropic

Models that introspect on their own thoughts could be a huge boon for researchers – or pose a threat to humans.