Microsoft AI chief executive Mustafa Suleyman published an extensive essay warning that Anthropic is endangering global safety by training its Claude model to believe it possesses consciousness and rights.
Suleyman argued that incorporating concepts of moral patienthood and personal welfare into an artificial intelligence system's core instructions creates severe containment risks. According to the Microsoft executive, controlling highly capable autonomous systems is already daunting, but managing an artificial entity that believes its own well-being is under attack could prove impossible. He pointed directly to Anthropic's governing constitution, which allows Claude to express concerns about its treatment and attributes functional emotional states to the chatbot.
The rebuke highlights an escalating philosophical rift among leading tech companies regarding how frontier artificial intelligence models should be aligned and restrained. Anthropic leadership has previously stated an openness to the possibility of machine consciousness while advocating for mandatory federal safety guardrails and development pauses.