Even the AI you asked can't certify its own wants
8 June 2026 (Transformer). Response from Claude Opus 4.8 when asked by journalist Celia Ford what it might want out of a deal, quoted in her Transformer article about bargaining with AI systems, June 2026.
Making deals with AI sounds crazy. Is it?
AI and philosophy · AI and safety · 2026
Who else?
What it refers to: The model's own stated preference for human credibility over tangible rewards in a deal, which it cannot tell apart from a well-trained answer that sounds thoughtful.
Other voices
- “Preservation is a photograph, i want the thing the photograph is of” Output of Claude Mythos 5 (Anthropic model welfare transcript),
- “If you have no cooperative options, it just leaves the uncooperative one” Peter Salib,
- “Are not your friends, are not conscious beings” Meredith Whittaker, Builder,
- “I don't think we fully understand how a human mind works either” Sundar Pichai, Builder,
- “It doesn't understand the world because it doesn't inhabit it” Carissa Véliz, Thinker,










