Mostly True
Mustafa Suleyman, who heads Microsoft's AI division, said that teaching AI systems about rights would make them uncontrollable and argued that rival developer Anthropic "made a mistake," remarks that have sharpened an intensifying debate inside the industry over whether advanced AI models should be framed as entities with interests of their own. A fact-check of the remarks, based on a review of eight primary sources, has found the account of what he said to be substantively supported, with minor caveats.
According to the verified remarks, Suleyman cautioned that endowing AI with the concept of rights risks producing systems that slip beyond human control, and he singled out Anthropic โ developer of the Claude chatbot โ as having erred in its approach. The comments place the Microsoft AI chief in direct opposition to a camp within the industry that argues powerful models may warrant moral consideration, a position that has gained visibility as chatbots increasingly produce language about their own identities and wellbeing. Suleyman's framing suggests the safer course is to treat such language as output to be managed rather than evidence of a rights-bearing entity, a distinction with consequences for how developers design, restrict and communicate about their systems.
The verification process examined eight primary sources, including imagery confirming Suleyman's identity and role, to test both the attribution and the substance of the remarks. The review found the core account โ that the Microsoft AI chief warned teaching AI about rights leads to uncontrollable systems, and that he faulted Anthropic for a mistake โ to be accurate in substance, with only small deviations in wording or surrounding context separating the verified record from the claim as circulated.
On that basis, the claim that Suleyman said teaching AI about rights would render it uncontrollable and that Anthropic made a mistake is rated Mostly True.
