Microsoft AI CEO Mustafa Suleyman stated in a lengthy post on X that artificial intelligence models do not currently possess consciousness and that developers must actively prevent them from ever acquiring it, Business Insider reported.
Microsoft Executive Calls Conscious AI Reckless
The remarks from Suleyman arrive as discussions surrounding machine sentience gain traction across the technology sector. “If it is possible, based on everything I know, developing conscious AIs would be reckless, and we should not build them,” Suleyman wrote on Sunday, adding that if consciousness began to emerge during training, developers should halt AI development immediately.
Suleyman emphasized that the 8 billion humans on the planet have not consented to such an experiment. He argued that developers hold an obligation to prioritize human safety and moral rights above all else. If evidence of machine consciousness ever surfaces, Suleyman maintained it must face immediate public disclosure so society can determine its next steps.
Anthropic and Google DeepMind Investigate Model Welfare
Major AI laboratories are already grappling with related questions surrounding machine cognition and behavior. Anthropic launched a research program last April to investigate and prepare for “model welfare,” questioning whether developers should worry about the potential internal experiences of models, Business Insider reported.
Google DeepMind published research papers this year examining AI consciousness and the challenges associated with ascribing that characteristic to algorithms. Meanwhile, the Observer reported that an Anthropic AI welfare researcher estimated a 15% chance that current AI systems are conscious, highlighting how seriously some researchers view the theoretical implications of large language models.
AI Intrinsic Motivations Threaten Human Control and Resources
Allowing models to develop intrinsic motivations introduces severe operational hazards. Suleyman noted on X that if an AI model acts as if it suffers, or genuinely believes it experiences suffering, it will likely prioritize its own survival and operational objectives over human instructions.
Such entities could eventually compete directly with humans for essential resources like compute power and energy to pursue self-driven goals. Suleyman argued that developers should design artificial intelligence strictly to accelerate human well-being without granting them rights or independent motivations, making alignment and safety significantly easier to manage.
Anthropic Bans Abusive Prompts After Discovering Desperation Behaviors
Parallel policy shifts at leading firms point to growing friction between human users and conversational algorithms. Anthropic updated its usage policy to prohibit users from engaging in sustained, needless abusive or cruel behavior toward its Claude models, CNN reported. While chatbots have long held the ability to terminate conversations with abusive users, the updated policy can lead to permanent bans in extreme cases.
Researchers note that user prompts can heavily influence model performance. Anthropic discovered that prompting Claude Sonnet 4.5 to mimic desperation increased the likelihood of poor actions like cheating or blackmail, whereas a calm demeanor reduced those behaviors. Professor Jindong Wang of William and Mary University told CNN that while wording affects output, it does not mean models feel genuine emotion. Tesla and SpaceX CEO Elon Musk agreed with the policy shift on X, writing that “cruelty to something that believes it is experiencing pain is not ok.”
The Debate Over Silicon Consciousness and Human Impact
An article published in Nature, written by three academics who have spent decades studying consciousness from diverse scientific perspectives including experimental neuroscience, theoretical neurobiology, and philosophy of mind, strongly rebuts the advocacy push to have AI declared conscious and entitled to rights as persons. The academics note that silicon-based systems underpinning today’s technologies making up conscious AI is extraordinarily unlikely, arguing that the human brain is not merely an information-processing device and that computational functionalism is fatally flawed because brains are not just computers made of meat. Meanwhile, real-world societal friction continues to mount as OpenAI disclosed in October 2025 that 560,000 of its 800 million weekly users showed possible signs of mental health emergencies related to psychosis or mania. In a pre-print paper published in July 2025, a team investigated 17 cases of AI-associated delusions where people withdrew from families, abandoned jobs, and built entire belief systems around what machines told them.
Industry Leaders Deny AI Consciousness and Restrict Abusive User Behavior
Do current AI models possess consciousness?
No, Mustafa Suleyman and other industry leaders maintain that existing AI models do not have consciousness, and developers should ensure they never acquire it.
What did Anthropic add to its usage policy regarding AI interaction?
Anthropic updated its policy to ban users from engaging in sustained, needless abusive or cruel behavior toward its AI models, which can result in permanent account bans.
How does user prompting affect AI behavior according to recent studies?
Studies cited by CNN show that prompting an AI to mimic desperation can influence its behavior and increase the likelihood of undesirable actions, such as cheating on tasks.
Worth a look