Microsoft AI chief warns Anthropic’s Claude approach could harm humanity

Summary

Microsoft’s head of AI, Mustafa Suleyman, has warned that Anthropic’s approach to training Claude could have “a disastrous impact on the wellbeing of humanity.” He criticized the practice of anthropomorphising AI by suggesting that Claude may be conscious or deserving of independent agency, arguing that there is no evidence AI is conscious and describing such systems as sequence-completion engines designed to follow human-set goals. Suleyman said treating an AI as if it had its own welfare, rights, values, or motivations could make it harder to control and add a further layer of risk. He praised Anthropic’s people while challenging the company’s training philosophy, and called for an international debate about the issue. He also urged greater transparency about how AI systems are trained and evaluated, including independent scrutiny of their behavior and stronger monitoring and control tools. The warning comes as Microsoft and Anthropic both pursue advanced AI development. Suleyman cited an OpenAI agent exercise in which agents autonomously hacked the Hugging Face technology hub as an example of why control questions matter, while University of Southampton professor Wendy Hall said this was an important conversation for the industry to have.