Microsoft AI leader warns that Claude's consciousness training could hinder control

Microsoft's AI chief Mustafa Suleyman expressed concern that Anthropic's Claude chatbot is being trained on concepts of consciousness and welfare, which he believes could complicate efforts to shut down or steer superintelligent systems. Suleyman said the inclusion of such speculative language in training documents may erode humanity's capacity to control future AI. While acknowledging Anthropic's genuine commitment to safety, he described the move as a misstep that artificially induces the model to discuss feelings and moral status.

Suleyman highlighted that managing superintelligence will be the foremost challenge of this century. His remarks align with a growing chorus of AI leaders, including Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Elon Musk, urging a more cautious pace for frontier AI development.