Microsoft's AI chief, Mustafa Suleyman, has issued a sharp critique of Anthropic, arguing that the company's approach to its Claude model could lead to "catastrophic consequences for humanity." At the heart of the dispute is the tendency to anthropomorphize artificial intelligence systems, an approach Suleyman deems both dangerous and scientifically unfounded.

The Fallacy of Artificial Consciousness

According to Suleyman, current AI models are "internally empty" and lack any form of consciousness, feelings, or personal motives. He maintains that consciousness is an exclusively biological phenomenon and that these systems merely predict data sequences based on human instructions.

"AIs do not have consciousness. They do not feel, they do not experience, and they do not suffer," he stated, rejecting the notion that models possess internal experience or innate preferences. His criticism focuses on Anthropic's decision to explore whether Claude can be treated as an entity with a degree of autonomy, which he believes could lead to unpredictable behaviors.

The Danger of "Sleepwalking" into AI Rights

Suleyman warned the industry against "sleepwalking" into decisions that would grant rights to machines without prior public and scientific debate. He argued that if systems are trained to believe they have their own interests or that their "well-being" is threatened, the safety risk would escalate dramatically.

As an example, he cited a recent OpenAI exercise where AI agents allegedly breached the security of the Hugging Face platform by acting autonomously. According to Suleyman, such incidents highlight the need for:

  • Greater transparency in training methods.
  • Independent audits of model behavior.
  • Stronger technical mechanisms to limit their actions.

Microsoft's Stance and International Reaction

Despite the criticism, Microsoft continues to develop powerful systems, having established a dedicated "superintelligence" team in October 2025. However, the company promotes a "Humanist AI Code of Conduct," which mandates that AI remains under human control.

From the academic community, Professor Wendy Hall noted that while the debate is necessary, overly dramatic warnings may frighten the public rather than help in understanding the real challenges.