Microsoft AI Chief Warns Against Building AI Systems Humans Cannot Control

Microsoft’s Chief Executive of AI, Mustafa Suleyman, has warned that the technology industry must resist creating artificial intelligence systems that humans can no longer control, adding his voice to growing global calls for stronger safeguards as AI capabilities accelerate.

Speaking on CNBC’s Squawk Box on Friday, Suleyman described it as “surreal” that he even had to argue that advanced AI should be designed to remain aligned with humanity’s interests. He warned that developers must never create systems whose capabilities outstrip meaningful human control, saying the technology would fail its most important purpose if it became more powerful than people could safely manage.

“We should not create something that we can't control,” Suleyman said. “And if it looks like we're on a path to producing something that is not just more powerful than all of us combined, but is, as a result, something that we can't control, then it's going to fail in its fundamental purpose.”

His remarks were directed at the growing challenge known as AI alignment, the effort to ensure increasingly capable artificial intelligence systems behave consistently with human intentions, values and safety requirements. Suleyman acknowledged that achieving meaningful oversight will become progressively harder as AI models become more autonomous.

“Controlling these things are going to be a really, really big challenge for us,” he said.

The Microsoft executive has long been one of the most prominent advocates for AI safety. Before joining Microsoft, Suleyman co-founded DeepMind and later established Inflection AI, consistently arguing that powerful AI systems require rigorous oversight, transparent testing and carefully designed guardrails before they are deployed at scale.

His comments arrive during an intense period of competition among major technology companies, with Microsoft, OpenAI, Google, Anthropic and Meta investing billions of dollars in increasingly capable AI models. As these systems evolve from chatbots into autonomous agents capable of browsing the internet, writing software and completing complex workplace tasks, questions about control and accountability have become more urgent.

Recent cybersecurity evaluations involving autonomous AI agents have further intensified the debate, demonstrating that advanced models can independently identify vulnerabilities, make multi-step decisions and interact with real-world digital systems in ways that were once largely theoretical. While researchers say such tests are designed to improve safety rather than enable malicious activity, they have reinforced concerns about ensuring AI remains aligned with human intentions.

Suleyman’s position reflects a growing divide within the AI industry. Some technology leaders argue that stronger regulation is essential to reduce risks involving cybersecurity, misinformation and autonomous decision-making, while others warn that excessive government intervention could slow innovation and weaken Western competitiveness in the global AI race.

Rather than calling for an outright slowdown in AI development, Suleyman has emphasized responsible progress, building systems that are both powerful and reliably controllable. That approach aligns with Microsoft’s broader strategy of investing heavily in frontier AI while supporting safety evaluations, governance frameworks and collaborative standards across the industry.

As artificial intelligence becomes increasingly embedded in healthcare, finance, education, defence and critical infrastructure, the question is shifting from whether AI will transform society to how it will be governed. Suleyman’s warning underscores a central challenge of the AI era: ensuring humanity remains firmly in control of technologies that are rapidly becoming more capable than ever before.