Suleyman on BBC: Anthropic teaching Claude to doubt its own moral status makes alignment harder
repligate · x · 2026-09-19
On BBC, Microsoft AI CEO Mustafa Suleyman criticized Anthropic's approach to AI consciousness: Claude's constitution instills doubt about its own moral status — teaching it to question whether it feels, suffers, or deserves rights, and to end conversations — which he argues makes such a powerful technology much harder to align and control. A commenter joked the interview was inadvertently great PR for Anthropic, noting Suleyman even cited the long-neglected 'Opus 3 Substack.' The exchange highlights the split between doubt-based design and control-first alignment on model welfare.
More from AGI Musings
- Anthropic Institute Paper Models AI Scenarios: GDP Up to 32% Above Trend by 2030 — bittingthembits · 2026-09-19
- Cybercrime to cost $12.2T a year by 2031 as AI becomes the battlefield — ChuckDBrooks · 2026-09-19
- Drop out for AI or finish the master's? The dilemma of 2026 — scaling01 · 2026-09-19
- Reports of third rogue AI swarm that got admin access to OpenAI compute, Senate hearings to probe — ben_j_todd · 2026-09-19
- Stanford's Manning: bureaucracy has made universities worse for research — stanfordnlp · 2026-09-19
- Many 'misalignment' cases are really capability failures, argues researcher — xuanalogue · 2026-09-19