Microsoft AI chief Suleyman attacks Anthropic's 'dangerous' consciousness claims while racing to displace its models

Mustafa Suleyman, CEO of Microsoft AI, has repeatedly warned that treating artificial intelligence as conscious is actively harmful. He told CNBC that only biological beings are capable of consciousness and that developers should stop pursuing projects suggesting otherwise. In a Decoder podcast interview with The Verge, he singled out Anthropic for speculating about Claude's consciousness in its constitutional documents, calling this practice really, really dangerous. Suleyman's concern is not merely semantic. He has linked seemingly conscious AI to rising reports of what he terms AI psychosis, where users perceive sentience in machines that lack it. On X, he wrote that such perceptions keep him awake at night and carry real societal impact even though, as he put it, there is zero evidence of AI consciousness today.

Suleyman's position has hardened over time. Wired reported in 2025 that he called machine consciousness an illusion and designing systems to mimic it dangerous and misguided. His stance now carries the weight of his role running Microsoft's in-house AI division, making it a direct challenge to competitors who employ more anthropomorphic marketing.

Suleyman's criticism of Anthropic's consciousness framing coincides with an intensifying competitive focus. At Microsoft's Build developer conference, he revealed that his team is more concerned about Anthropic than Google, Meta, or even OpenAI, Microsoft's closest partner. The reasoning is structural. Anthropic's aggressive push into enterprise software and coding tools threatens Microsoft's core corporate business directly. Suleyman told the Financial Times that Microsoft is more focused on the Anthropic-style which is enterprise use cases, developers and coding.

This represents a notable shift in competitive framing. While public attention fixates on OpenAI versus Google, Microsoft's internal strategic calculations center on Anthropic's encroachment on the business productivity and developer tooling markets that generate the bulk of Microsoft's revenue. The consciousness debate, in this light, doubles as a way to differentiate Microsoft's more restrained product philosophy from Anthropic's more expansive claims about its models.

Behind the philosophical disagreements lies a concrete financial battle. Suleyman told Bloomberg that Anthropic is extremely expensive and that many people are urgently looking for alternatives. He was explicit about Microsoft's own motivation: We pay a lot of money to Anthropic, so our goal is to reduce and ultimately eliminate that cost. Microsoft is reportedly cancelling subscriptions to Claude Code for many of its engineers as part of this push.

To accelerate independence, Microsoft unveiled seven in-house AI models at Build 2026 designed to compete on price. The timing is strategic, as Microsoft also introduces usage-based pricing for GitHub Copilot that shifts cost burdens to developers. Suleyman's public cost complaints serve dual purposes, pressuring Anthropic on pricing while justifying Microsoft's own pricing changes to customers. The models span coding, enterprise search, and document analysis, directly targeting Anthropic's strongest product categories.

Suleyman's consciousness critique functions as competitive positioning with ethical overtones. By framing Anthropic's approach as dangerous, he attempts to seize the responsible AI mantle while undermining a rival's brand differentiation. Anthropic has built significant market presence partly on claims of more careful, constitution-guided AI development. Suleyman's attack strikes at that foundation.

The strategy also distances Microsoft from its own partnership with OpenAI, which has faced criticism for similar anthropomorphic tendencies in ChatGPT's design. Suleyman can present Microsoft as the sober alternative, even as the company deeply integrates AI across Windows, Office, and Azure. Harvard Law Review has documented how this competitive dynamic, where safety claims and commercial interests interweave, creates governance challenges for the AI industry. Suleyman's personal credibility, built through DeepMind cofounding and Inflection's consumer AI work, lends his arguments more weight than typical corporate messaging.

Microsoft's seven-model rollout signals an acceleration of its in-house AI capabilities, reducing dependency on both OpenAI and Anthropic. Suleyman's comments suggest near-term pressure on Anthropic to defend its pricing or risk losing enterprise customers to cheaper alternatives. The consciousness debate will likely intensify as models become more capable, with each company seeking to define the terms of acceptable AI behavior.

For developers and enterprise buyers, the competition promises more options and potentially lower costs, but also more complexity in evaluating claims about AI capabilities and limitations. Suleyman's intervention ensures that the consciousness question will remain contested terrain, not merely a technical debate but a strategic one with real commercial consequences.