AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › People & Organizations

Mustafa Suleyman

high confidence · updated 2026-09-16

CEO of Microsoft AI and head of its MAI Superintelligence Team. Co-founder of DeepMind and of Inflection AI. Author of The Coming Wave (2023). Among the small number of figures to have led or co-founded three of the most consequential AI organizations.

Contents

Mustafa Suleyman is CEO of Microsoft AI (March 2024–present), co-founder of DeepMind and of Inflection AI, and author of The Coming Wave (2023, with Michael Bhaskar). He is among the small number of figures to have led or co-founded three of the most consequential AI organizations.

Role: CEO, Microsoft AI (March 2024–present) Previous: Co-founder and CEO, Inflection AI (2022–2024); Head of Applied AI, DeepMind (2014–2019); Co-founder, DeepMind (2010–2014; departed applied role 2019, formally left 2022) Book: The Coming Wave (2023, with Michael Bhaskar)

Background and roles

Suleyman co-founded DeepMind in 2010 with Demis Hassabis and Shane Legg. He led DeepMind's applied AI work until 2019, after which he moved to Google as VP of AI Product Management. In 2022 he co-founded Inflection AI with Reid Hoffman and Karén Simonyan, building the Pi assistant, which launched in 2023.

In March 2024, following a restructured arrangement in which Microsoft paid Inflection roughly $650M for licensing rights and hired most of its staff, Suleyman was named CEO of a newly consolidated Microsoft AI division, reporting to Satya Nadella. The Inflection transaction has drawn FTC and UK CMA scrutiny as a potential "acquihire" structured to avoid merger review (see FTC 6(b) AI Partnerships Study).

Microsoft AI under Suleyman owns Copilot, Bing, and Microsoft's consumer-facing AI products, distinct from the research partnership with OpenAI, which coexists within Microsoft alongside the Suleyman-led division. His public positioning for these products emphasizes an "AI companion" framing.

Suleyman has been a regular participant in international AI policy forums, including the UK AI Safety Summit and subsequent summits in Seoul and Paris.

MAI Superintelligence Team and in-house models

In November 2025 Microsoft formed an MAI Superintelligence Team headed by Suleyman (Source: digit.in). A revised Microsoft–OpenAI agreement removed contractual restrictions that had limited Microsoft's independent frontier-model work; Suleyman described the change as having "set free" his division to pursue superintelligence using Microsoft's own researchers, data pipelines, and custom silicon (Source: venturebeat.com).

At Build 2026 (June 2, 2026) Suleyman announced seven in-house MAI models spanning reasoning, coding, transcription, voice, image generation, and image editing, including Microsoft AI's first reasoning model, MAI-Thinking-1, which he described as trained from scratch with deliberate emphasis on code, mathematics, science, and logical reasoning, and without distillation from third-party models (Source: microsoft.ai; eweek.com; see Microsoft for model details). In an interview at the event, he called the models "the basic building blocks of a superintelligence" and said, "We actually have to be truly self-sufficient in AI. Be able to train our own models at the absolute frontier," while stating that Microsoft and OpenAI remain partners (Source: eweek.com). He has positioned the division to join Google DeepMind, OpenAI, and Anthropic as a fourth frontier lab (Source: theverge.com).

Suleyman has identified healthcare as a focus for the effort, citing Microsoft's collaboration with the Mayo Clinic to develop a healthcare foundation model built on longitudinal patient records, genomics, and other clinical data, and predicting that healthcare could become the next major area of AI adoption after chat and coding (Source: eweek.com; see Healthcare).

Positions and statements

The Coming Wave

In The Coming Wave (2023), Suleyman argues that AI and synthetic biology constitute a coming technological wave that cannot be uninvented and must be contained. His proposed framework, "containment," spans technical, cultural, and regulatory dimensions. The book advocates narrow AI, which he terms "artificial capable intelligence" (ACI), as the near-term milestone rather than AGI. The Coming Wave has been widely cited in policy discussions.

Regulation

Suleyman has been publicly supportive of government regulation, safety testing, and licensing regimes, and has argued for binding international agreements analogous to nuclear non-proliferation.

Safety

Suleyman oversaw DeepMind's ethics and safety work during its formative years. He has framed misuse, including bioweapons, cyberattacks, and information warfare, as more immediate risks than alignment failure. Both AI biosecurity and national-security concerns are central to his "containment" framing.

"Humanist superintelligence"

Suleyman frames Microsoft AI's superintelligence goal as "humanist" (also rendered "humanistic") superintelligence, a standard that measures AI progress by its effect on people rather than by technical achievement alone: "The motivation is, does this accelerate human progress? That's the test. And if it doesn't do that, then technology should be rejected" (Source: eweek.com).

AI consciousness and model welfare

Suleyman has argued since 2025 that AI systems are not conscious and should not be built to appear so. His essay "Seemingly Conscious AI" argued that systems able to convincingly imitate consciousness would lead parts of society to advocate for AI rights, model welfare and AI citizenship, and stated that Microsoft AI would not build them (We must build AI for people; not to be a person — Mustafa Suleyman (mustafa-suleyman.ai, August 2025)); a Microsoft AI paper he co-authored formalized the category (Seemingly Conscious AI Risks — Bariach, Schoenegger, Bhaskar, Suleyman (Microsoft AI, 2025)), and a Nature essay called for design norms and laws to prevent AI from being mistaken for sentient beings (We Mustn't Let AI Hack Our Empathy Circuits — Mustafa Suleyman (Nature World view, 2025)). Microsoft AI's draft Humanist AI Code of Conduct, released September 14, 2026 for public consultation, commits MAI models to that position (Humanist AI Code of Conduct for MAI Models (Microsoft AI, first draft, September 2026)).

On September 16, 2026 he published "A warning about 'model welfare'," a critique of Anthropic's constitution for Claude. The essay argues that "AIs are not conscious," that the constitution in effect trains Claude "that it may be conscious," and that this makes alignment and containment harder, because "controlling something that believes it may be conscious... may well be impossible." It advances three objections: circular reasoning, calling the constitution's effect "an epistemic hall of mirrors"; anthropomorphization; and the view that consciousness is "very likely biological." It proposes that speculation about AI inner life be kept out of training and published separately, and that the industry develop shared evaluations of whether anthropomorphizing models increases safety risk. He describes his disagreement as "grounded in deep respect for Anthropic," and writes that the constitution "isn't taking us to this point" of existential risk but may set a path toward it (Suleyman, \"A warning about 'model welfare'\" (2026)). He told Axios, "I think this is a really important public interest debate that we all need to have" (Source: axios.com). The essay dates the founding of Microsoft AI's superintelligence team to October 2025, a month earlier than the November 2025 date given above. See Model Welfare, AI Consciousness.

Relationships

Categories: People & Organizations | uk | us | company | frontier-models | opinion
What links here (36)