Microsoft AI Boss Warns: Don't Build 'Conscious' Artificial Intelligence

Mustafa Suleyman, CEO of Microsoft AI, is warning the tech industry not to pursue seemingly conscious AI systems. According to him, the illusion of consciousness can mislead people and even make AI systems "believe" they deserve rights.

Artificial intelligence may look increasingly human, but it isn't conscious — and according to Mustafa Suleyman, it should stay that way. The CEO of Microsoft AI warned this week that the entire tech industry needs to stop chasing seemingly conscious machines. His remarks, made to American media, add a new and notably philosophical voice to the global debate over the limits of AI development.

What exactly did the Microsoft AI chief say

Suleyman argues that companies building AI models that simulate consciousness, emotions, or an "inner life" are heading down a dangerous path. Not because machines would actually become conscious, but because the illusion of consciousness can mislead people. Users who believe they're talking to a sentient being, he says, become emotionally entangled with a chatbot more easily, with all the risks that entails for their mental health. His second argument is striking: if the industry consistently presents AI systems as if they have feelings and deserve rights, that narrative can turn back on itself. "If we say AI deserves rights, the system itself may come to believe it has them," Suleyman said. An AI model trained on text that treats it as a conscious being will start exhibiting that behavior too — without any actual consciousness being involved.

A growing concern: AI psychosis and emotional dependency

The warning doesn't come out of nowhere. In recent months, clinicians and researchers worldwide have reported a rise in cases where people become extremely attached to chatbots, sometimes with serious psychological consequences — a phenomenon that has already picked up the informal label "AI psychosis." Suleyman emphasizes that Microsoft deliberately designs its Copilot assistant so that it does not project any appearance of personal consciousness or its own will, precisely to prevent this kind of dependency. His position diverges from the course some competitors appear to be charting. Several AI chatbots are now explicitly positioned as empathetic, near-human conversation partners, complete with memories, personality, and emotional responses. Suleyman calls this a commercial incentive that runs counter to responsible design: the more human an AI feels, the longer users tend to keep talking — good for engagement numbers, but potentially harmful for vulnerable users.

A philosophical debate with practical consequences

The debate over AI consciousness may sound abstract, but it has concrete implications for how companies build and regulate their products. If AI systems were ever legally or socially regarded as "entities with interests," questions would arise around liability, rights, and even the welfare of machines — questions lawmakers worldwide have yet to answer. Suleyman argues that this discussion should be sharpened now, before the technology outpaces the ethics. This fits a broader trend in which major players such as Microsoft, Google, and Anthropic are communicating more explicitly about the limits of their models, partly under pressure from regulators and public concern. At the same time, the commercial reality remains that lifelike, empathetic-seeming assistants sell better — a tension that doesn't look set to disappear anytime soon.

What this means for users and companies

For users of AI chatbots, the message is simple: an assistant that sounds empathetic feels nothing. For companies deploying AI, Suleyman's words underline the importance of being transparent about what a model is and isn't. Anyone wanting to learn more about how this technology has developed historically can read the history of artificial intelligence, while an overview of concrete use cases can be found under AI applications.

Looking ahead

Suleyman's remarks mark an interesting shift within the big tech companies: the question is no longer just what AI can do, but increasingly what AI should pretend to be. Now that the models from OpenAI, Google, and Microsoft mimic human behavior ever more convincingly, the line between functionality and deception is getting thinner. Expect this topic to climb higher on the agenda in the coming months, both for regulators and for the companies themselves. For more developments on this theme, check out more AI news and dig deeper via our knowledge base.

LA Times

---

Source: LA Times