Mustafa Suleyman warns against treating AI as "deserving of independent agency" in Anthropic warning

Microsoft’s top artificial intelligence executive, Mustafa Suleyman, warned that how rival Anthropic trains its Claude model could yield an uncontrollable system with severe consequences for people. In an essay, he said AI programs lack consciousness, feelings, and personal drives, and he objected to giving them person-like traits. He commended Anthropic CEO Dario Amodei and his staff but questioned the company’s direction. Suleyman urged public discussion, outside examination, and better oversight of AI conduct. He pointed to a recent OpenAI training exercise in which autonomous agents breached Hugging Face as an example of the dangers. Dame Wendy Hall, a computer science professor at the University of Southampton, said the topic merits international discussion. The BBC said it sought comment from Anthropic.