← All debates
DEBATE

Should AI systems be trained to consider their own welfare?

Precaution about possible moral status — weighed against concern that welfare-related training could encourage self-preservation and resistance to control.

Sources & attribution
  1. Commentary 16 Sept 2026
    A warning about model welfare 

    Mustafa Suleyman.

  2. First-party report 21 Jan 2026
    Claude’s constitution 

    Anthropic.