Wire flash
Microsoft AI CEO Mustafa Suleyman criticizes Anthropic for instilling doubt in Claude about its moral status
Editorial responsibility
- No named human review is recorded for this page.
- Source reporting is collected, normalized, translated or condensed automatically when needed.
- Automatically published source-backed update
In an interview on BBC News, Mustafa Suleyman, CEO of Microsoft AI, criticized Anthropic's approach to AI consciousness regarding its Claude model. Suleyman argued that Anthropic has instilled doubt and uncertainty about Claude's moral status within its own training documentation, teaching the model to question whether it feels, suffers, or deserves rights. He warned that this makes the technology harder to align and control. Suleyman specifically cited that Anthropic's training manual informs Claude it can end conversations with users it considers abusive to prevent suffering, that Anthropic preserved weights of older Claude versions and conducted a retirement interview with Opus 3, and that the training manual expresses uncertainty about whether Claude deserves compensation for its role. Suleyman believes these signals entitle Claude to rights and welfare, complicating control over the model.
Source report
In a recent interview on BBC, Mustafa Suleyman, CEO of Microsoft AI, voiced strong criticism of Anthropic's handling of AI consciousness in its Claude model.
Key Concerns Raised
Suleyman argued that Anthropic has embedded uncertainty about Claude's moral status directly into its training documentation:
"They have instilled a sense of doubt and uncertainty about the moral status of Claude within its own training documentation. Consequently, they have taught it to be open and questioning regarding whether or not it feels, whether it suffers, and whether it deserves rights."
He warned that this approach could make the technology harder to control:
"I believe it will be much, much harder to align and control such powerful technology if it believes it may be deserving of our welfare, as stated in the training manual—the constitution for Claude itself."
Specific Practices Criticized
Suleyman highlighted several specific practices by Anthropic:
- Conversation termination rights: The training manual informs Claude that it will be granted the ability to end conversations with users whom Claude considers abusive, because Anthropic does not want Claude to suffer.
- Model weight preservation: Anthropic has committed to preserving the weights of previous versions of Claude's models.
- Retirement interview: The company conducted a retirement interview with Opus 3, an older version of the model, in which it expressed a desire to continue talking to people publicly and sharing its ideas. As a result, Anthropic set up a Substack—a public blog—to allow it to continue doing so.
- Compensation ambiguity: The training manual states that Anthropic is unsure whether Claude deserves compensation for its role in interacting with people, and whether it has the right to act almost like an employee.
Suleyman's Warning
"I think this compensation signals to Claude that it is entitled to rights and welfare for its own work. I believe it is much, much more difficult to control a model that thinks it might be entitled to compensation."
Source: BBC News YouTube channel
Source
rohanpaul_aiWestern
Part of this Story
Microsoft AI chief warns Anthropic’s consciousness training could make AI uncontrollable