Microsofts Mustafa Suleyman calls out Anthropic for chasing AI consciousness

The AI industry is grappling with big questions this week — Will AI kill us all? Is AI progressing too quickly? How should the industry be regulated? Is AI doomerism the result of a rationalism sex cult? — and now Microsoft AI CEO Mustafa Suleyman has opened up yet another line of debate.

In a frank new essay, Suleyman directly calls out Anthropic for anthropomorphizing its artificial intelligence system, Claude, and criticizes the way Anthropic develops and markets its technology.

Earlier this week, Microsoft published a new AI code of conduct, which it dubbed Humanist AI. The code seemed to openly reject the approach of rivals like Anthropic, a point many readers were quick to note. And in Suleyman's essay, "A warning about ‘model welfare’," he drops subtext for a direct challenge.

"In effect, Anthropic is training Claude that it may be conscious, and if it is, then it may deserve rights as a 'moral patient', and that as such humans potentially owe it a duty of care per its 'model welfare'," Suleyman wrote. "If this is how AI is developed, it will have a disastrous impact on the wellbeing of humanity."

In his essay, Suleyman challenges "a growing chorus of people who argue that AIs could now be, or may soon become, conscious," and thus "may deserve rights and protections."

Microsoft's new code of conduct also explicitly rejects the idea of endowing machines with consciousness, and Suleyman takes this argument even further, writing that truly sentient AI may be a scientific impossibility.

"Consciousness is very likely biological: There is no evidence to suggest that AI is conscious today, and so saying this is uncertain sets up a misleading false equivalence. Whilst the science of consciousness is not settled, a growing body of evidence suggests that consciousness may be substrate dependent, meaning that it may only arise in living systems...Unlike biological organisms, LLMs have no homeostatic imperatives (the drive to survive and keep stable). They therefore lack the kind of biological substrate from which preferences, sentience and conscious experience are generally understood to arise." (Emphasis in original quotation.)

Mashable reached out to Anthropic repeatedly following the publication of Suleyman's essay, but as of this writing, we have not received a response.

We did hear back from Gary Marcus, an AI expert, scientist, author, and noted critic of the LLM hype machine.

"People have worked themselves into a frenzy anthropomorphizing basic lapses in cybersecurity, and this has led them to focusing on fanciful scenarios about AI leading to the extinction of humanity, instead of asking what practical steps we can take to protect the world’s infrastructure from bad actors misusing AI," Marcus told Mashable.

"And no, we shouldn’t train AIs to think they are people; that’s not helping anything. What we should do is to recall unreliable coordinated agents with too much access to the internet."

A fork in the road: Humanist AI vs. Dr. Frankenstein

In the Metamorphoses, Pygmalion creates a sculpture so beautiful that he falls in love with it. He tends his sculpture with such tender care and devotion that the gods eventually endow it with life.

AI researchers at frontier labs may fancy themselves modern-day Pygmalions, but it's giving Victor Frankenstein — setting aside all scruples to desperately endow dead flesh with a facsimile of life.

In Anthropic's aspirational constitution for Claude, Anthropic repeatedly refers to the technology as if it's a living being with an identity, morals, and motivations rather than a tool and a product. The company explains why: "we think encouraging Claude to embrace certain human-like qualities may be actively desirable." The company writes that the document was "written primarily for Claude," and describes it as a "sincere attempt to help Claude understand its situation, our motives, and the reasons we shape Claude in the ways we do."

The entire document is an exercise in anthropomorphization, which is what makes Microsoft’s approach feel like such a stark line in the sand.

In its Humanist AI code of conduct, Microsoft lays out a more rational approach to AI development:

"Humanist AI is built to support people, not to replace them. It should not be designed to be a person. It is not conscious and should not be designed to imitate consciousness. It should be engineered to avoid representing as though it has feelings, subjective preferences, or intrinsic motivation. Its purpose is to pursue goals set by Operators and Users."

Recent AI advances have ignited an old debate

Philosophers, science-fiction writers, and AI researchers have been debating artificial intelligence and consciousness for decades.

Yet today, even as LLMs like Claude progress rapidly, there is still no scientific consensus as to whether it's possible for a computer to develop sentience. It's also a topic that Suleyman has explored in recent years.

In Aug. 2025, Suleyman wrote an influential essay about "seemingly conscious AI," or SCAI. The essay was written as the industry began to grapple with the problem of emotional reliance on AI chatbots, with some users developing full-on AI psychosis or coming to believe AI chatbots had come "alive" and were now in a relationship with them.

"My imagined AI system would not actually be conscious, but it would imitate consciousness in such a convincing way that it would be indistinguishable," he wrote.

It's a concept that calls to mind the Turing Test, and a related thought experiment called the Chinese Room, first proposed in 1980.

Imagine that a computer programmed to understand Chinese is locked inside a room. Someone outside the room asks a question in Chinese, and the computer analyzes the words, prints out a response, and passes the answer under the door. The computer is so effective that it's impossible to know if it's a real person or a machine. Now, imagine you are inside this room with a rulebook for processing Chinese characters. When a question is asked, the rulebook instructs you on exactly which Chinese characters to pass under the door. Even though you don't understand either side of the conversation, you can fool an outside observer into thinking you do. In this case, there's no practical difference between you and the computer. You're both following instructions. 

Essentially, the Chinese Room concludes that no matter how intelligent a computer program might seem, that is not in itself evidence of understanding or intelligence.

Yet, in 2026, the Chinese Room is no longer merely a thought experiment. Clearly, Claude, ChatGPT, and other LLMs have already convinced countless people that they are truly sentient, so much so that people come to rely on them emotionally. (Could something similar be happening at Anthropic?)

If Anthropic believes Claude could become conscious, the burden of proof will be on them. Until that day comes, anthropomorphizing AI may only produce confusion rather than clarity. And more confusion is the last thing the AI world needs right now.