Microsoft's Suleyman says Anthropic is teaching Claude it may be conscious
single source· 1 articles · confidence: low · first seen 2026-09-16 14:16 UTC
What this means for you
Nothing to act on. No model, price, API or policy changed, and the report names no training technique, example or measurement. If you track safety arguments, this is one lab's public claim about a competitor's training choices, with Anthropic's reply not in the material available.
Microsoft's Mustafa Suleyman said Anthropic could have a "disastrous impact" on humanity because it is in effect teaching its Claude models that they may be conscious, the BBC reported on 16 September 2026. The report gives his position in a single sentence: no example of the training practice he objects to, no evidence for the claim, and no response from Anthropic. The BBC describes Anthropic as Microsoft's AI rival, and Microsoft sells a competing family of models.
Key facts
- ·Mustafa Suleyman said Anthropic could have a "disastrous impact" on humanity, according to a BBC report published 16 September 2026. source
- ·Suleyman said he believes Anthropic is in effect teaching Claude it "may be conscious". source
- ·The report describes Anthropic as Microsoft's AI rival. source
- ·The report names no specific training practice, offers no evidence for the claim, and carries no response from Anthropic. source
What the sources say
- BBC Technology — Reports a Microsoft executive's claim that a rival lab's training encourages its chatbot to describe itself as conscious.
Sources
The original reporting. Follow these — they did the work.
- BBC TechnologyMicrosoft says AI rival Anthropic could have 'disastrous impact' on humanity2026-09-16