Mustafa Suleyman, Microsoft’s AI chief for 18 months, detailed AI’s superhuman status and humanist superintelligence goals in a remote interview from Seattle with Mishal Husain, following a revised OpenAI deal amid surging data-center investments.
The AI race entered an uncharted and expensive phase in 2025, characterized by mega-deals involving billions of dollars invested in data centers, cross-investments among leading companies, and a talent arms race attracting top minds. These developments marked a year of intensified competition and resource allocation in artificial intelligence development.
Suleyman co-founded DeepMind, which Google acquired in 2014. DeepMind later developed an AI system that defeated a world champion at the game of Go, demonstrating advanced capabilities in complex strategic gameplay. At Microsoft, Suleyman’s initiatives faced limitations from an existing deal with OpenAI until a revised agreement permitted him to publicly outline new objectives.
The interview occurred remotely at an early hour in Seattle, as Suleyman’s team had anticipated his presence on the east coast. Despite the timing, Suleyman engaged directly, displaying enthusiasm at points while maintaining realism and occasionally revealing a political viewpoint uncommon in Big Tech discussions.
Suleyman described a personal use of AI through Copilot, a tool he employs to maintain a table recording films he loves, organized by date. He adds personal notes, and Copilot provides links to film posters. After watching a film late one night, he updated this table and prompted Copilot with requests such as, “What would be a similar one?” He stated, “It’s possible to ask your AI to do pretty much any knowledge work task — just like you might ask an assistant to organize your life. The more obscure, creative [and] challenging the task you’re going to ask your AI, the better.”
During the video call, Suleyman’s bookshelf in Seattle revealed his reading interests, including recent books by Michael Wolff and Robert Kaplan, as well as “The Tech Coup: How to Save Democracy from Silicon Valley” and “Gaza: An Inquest Into Its Martyrdom.” These titles indicated a range of political and contemporary topics.
On autonomous AI tasks, Suleyman noted ongoing experimentation. AI can book tickets or purchase gifts, though not always accurately. Such functions remain in “dev mode” and unavailable to the general public. When operational, the process involves AI typing into browsers, clicking buttons, opening tabs, and reviewing user history for personalization. He called it “the most magical thing you’ve ever seen.”
Addressing errors, Suleyman explained that AI might select the wrong item, such as buying a present for the incorrect person, but users can intervene. AI always seeks permission before proceeding, ensuring safety. He reflected, “It’s a funny thing, technology. It’s magical and amazing, but it’s always just got a little bit further to go. In this case, a while yet before it’s everyday.”
Drawing from his experience founding DeepMind and later Inflection before joining Microsoft, Suleyman expressed stoicism toward these setbacks. He anticipated reliable performance within six months, 12 months, or at most 18 months. He asserted, “It is already superhuman.”
Suleyman predicted that by the following Christmas, consumers could purchase presents using autonomous AI agents, deeming it “highly likely.”
The term superintelligence gained prominence in public discourse recently, influenced by Suleyman and others. In January 2025, Sam Altman announced OpenAI’s aim beyond artificial general intelligence (AGI, matching human capabilities) toward superintelligence. Philosopher Nick Bostrom popularized the concept. In June 2025, Mark Zuckerberg restructured Meta’s AI division into Meta Superintelligence Labs. Last month, Suleyman introduced Microsoft AI’s Superintelligence Team.
Suleyman defined superintelligence as an AI system capable of learning any new task and performing better than all humans combined across all tasks. This sets a high threshold accompanied by substantial risks, particularly uncertainty in containing and aligning such a powerful system with human interests. He advocated for “humanist superintelligence,” always aligned to human interests and safe. Microsoft commits to proving safety before advancing systems that could escape control. He emphasized, “Everybody should agree to that. Yet I think it’s a novel position in the industry at the moment.”
Microsoft positions itself through this humanist lens. Established 50 years ago, the company maintains a reputation for caution and trust, with 90% of S&P 500 companies relying on its email, operating systems, and productivity tools. This careful approach continues, with humanist superintelligence as a core element.
Industry observers note potential tensions between this stance and commercial pressures to justify AI investments. Historical precedents include OpenAI’s founding by Sam Altman and Elon Musk due to concerns over Google’s trustworthiness in AI leadership. In 2021, some OpenAI staff departed to form Anthropic over safety disagreements with OpenAI.
Regarding rivals like OpenAI, Meta, and Anthropic, Suleyman refrained from judgment, stating, “Everybody has to decide what they stand for and how they operate. I don’t want to judge how they’re operating right now.” He observed no evidence of large-scale harm, self-improving systems, or autonomous operations currently. Within five to ten years, capabilities like self-goal setting, code improvement, and autonomy may emerge, heightening risks. He called for caution through transparency, audits, government involvement, and proactive disclosures on proximity to these capabilities, describing it as “obvious.”
Suleyman confirmed Microsoft will not release superintelligence tools until confident in control, with containment and alignment as prerequisites and red lines. He urged industry-wide adoption, questioning, “Are they building a humanist superintelligence?” He added that despite disagreements, all parties prioritize species survival and human flourishing.
Confidence in controls remains challenging, as incidents occur even with humans. Microsoft disabled services used by the Israeli Ministry of Defense following a Guardian report alleging use for mass surveillance. Suleyman praised the reporting, noting, “We were very grateful for it. As soon as we became aware of it, we made all of the necessary changes and removed the IDF from those servers. They were clearly not in compliance with our terms of service and there’s an ongoing investigation internally.” Initially, Microsoft stated no evidence of harm or violations of terms, AI Code of Conduct, or compliance by the Israeli government. Currently, a data protection and legal complaint in Ireland, Microsoft’s European headquarters, alleges continued hosting of Palestinian-monitoring applications in its data centers.
Suleyman stressed auditing systems and swiftly removing violators of terms. These complex systems carry significant risks, but rapid response represents the optimal mitigation strategy.
The initial applications of superintelligence will occur in medicine. Existing systems diagnose any rare condition documented in literature, surpassing human performance in accuracy, cost, and required tests. Independent peer review is underway, followed by clinical trials. Suleyman described this as “very, very, very exciting.”
This medical focus aligns personally with Suleyman, whose mother worked as a nurse. He believes technology serves humanity by improving comfort, extending longevity, reducing work hours optionally, and generating abundance. He prioritizes such applications deliberately.
Suleyman defined abundance as the inevitable outcome where, within 20 to 30 years—or sooner—machines exceed humans at most work. Society must determine purpose, thoughtfully pace machine introduction, and implement funding and support mechanisms to offset worker displacement during transition.
He endorsed universal basic income as inevitable and desirable, a long-held position. Abundance already exists but distributes unevenly. Value extends beyond physical atoms like food and cars to digital elements such as ideas, knowledge, and intelligence, which proliferate rapidly. Large language models and chatbots achieved 2 billion annual users in three years, the fastest-spreading technology in history. Competitive pressures will lower AI access costs. The key challenge involves taxation and redistribution for a healthy transition.
Suleyman’s perspectives trace to a working-class upbringing. His father drove a cab, and the family led an unremarkable life. His parents undervalued formal education, encouraging a trade instead.
The interview extended into further details on Suleyman’s background, personal experiences, vision for AI, evolving ties with OpenAI and Microsoft, dynamics within the tech ecosystem, regulatory considerations, and prospective directions, as captured in the full transcript.
