Slow frontier development as superintelligence approaches so coordination, containment, alignment and independent auditing can catch up; continue bounded humanist AI, without a blanket halt. [1][2][3]
Initial statement. Research draft; independent adversarial review and human editorial approval remain separate.
Source [1] · Original statement
Microsoft AI CEO says AI threats are real, and Anthropic is making it worse
Attributed to: Mustafa Suleyman · The Verge
Published
“…it seems to me very straightforward that we would want to slow down”
Final section, answer immediately preceding Patel’s question “Is it fair to say that you are also calling for a slowdown?” · View original source ↗
Source context. Original Decoder edited transcript by Nilay Patel. The quoted clause is conditioned by “If we are approaching that point, superintelligence, in the next few years”. He subsequently answers “Yeah” to slowdown question, explains independent evaluators, and elsewhere rejects categorical stop-now or accelerate-only claims.
Scope
Frontier development as superintelligence approaches in the next few years.
Conditions
Coordinate; establish containment, alignment and regulatory auditing; broadly appointed independent embedded evaluators; no categorical immediate stop.
Stated rationale
Interpret only the verified passage in its stated scope; do not infer pace from identity, optimism, risk concern or deployment policy.
Source details
primary · Original primary HTML independently fetched/read after web tool blocked. JSON-LD publication2026-09-17T14:00:00Z; recording date not stated. Locator uses unique question/answer, no invented timestamp. Newer than November2025 bounded acceleration.
Discovered 3 Oct 2026.
Published 17 Sep 2026.
Accessed 3 Oct 2026.
Source [2] · Original statement
Towards Humanist Superintelligence
Attributed to: Mustafa Suleyman · Microsoft AI
Published
“It means proactively avoiding harm and then accelerating.”
Humanist superintelligence section, final sentence of paragraph beginning “Humanist superintelligence (HSI) offers an alternative vision” · View original source ↗
Source context. Authored by Mustafa Suleyman; he defines HSI and says human-centrism precedes acceleration. The essay rejects superintelligence at any cost.
Scope
Humanist superintelligence: highly capable, domain-specific systems serving humans; not unbounded autonomous superintelligence.
Conditions
Human interests first; avoid harm before acceleration; containment, alignment and human control.
Stated rationale
Interpret only the verified passage in its stated scope; do not infer pace from identity, optimism, risk concern or deployment policy.
Source details
primary · Independently opened original source; Grok is not evidence.
Discovered 3 Oct 2026.
Published 6 Nov 2025.
Accessed 3 Oct 2026.
Source [3] · Original statement
Announcing Copilot leadership update
Attributed to: Mustafa Suleyman · Microsoft
Statement
“This requires us to build frontier models, at scale, pushing the boundaries of what’s possible.”
Source context. Letter shared with employees that morning; separate Satya and Mustafa messages. This quote is in Mustafa’s own section.
Scope
Microsoft AI frontier-model and superintelligence development.
Conditions
Transformative positive impact; humanist mission.
Stated rationale
Interpret only the verified passage in its stated scope; do not infer pace from identity, optimism, risk concern or deployment policy.
Source details
primary · Independently opened original source; Grok is not evidence.
Discovered 3 Oct 2026.
Published 17 Mar 2026.
Accessed 3 Oct 2026.
Editorial assessment. November2025 personally advocated accelerating bounded humanist superintelligence after avoiding harm. September17 2026 explicitly supports conditional slowdown and rejects categorical stop-now or accelerate-only answers. Continued Microsoft frontier work is not erased. Unresolved: None. No human publication approval.
Assessment reviewed
Earlier position ·
ACCELERATE
Accelerate bounded humanist superintelligence after proactively avoiding harm; human interests come first. [1]
Scope distinction. Accelerate bounded humanist superintelligence after proactively avoiding harm; human interests come first.
Source [1] · Original statement
Towards Humanist Superintelligence
Attributed to: Mustafa Suleyman · Microsoft AI
Published
“It means proactively avoiding harm and then accelerating.”
Humanist superintelligence section, final sentence of paragraph beginning “Humanist superintelligence (HSI) offers an alternative vision” · View original source ↗
Source context. Authored by Mustafa Suleyman; he defines HSI and says human-centrism precedes acceleration. The essay rejects superintelligence at any cost.
Scope
Humanist superintelligence: highly capable, domain-specific systems serving humans; not unbounded autonomous superintelligence.
Conditions
Human interests first; avoid harm before acceleration; containment, alignment and human control.
Stated rationale
Interpret only the verified passage in its stated scope; do not infer pace from identity, optimism, risk concern or deployment policy.
Source details
primary · Independently opened original source; Grok is not evidence.
Discovered 3 Oct 2026.
Published 6 Nov 2025.
Accessed 3 Oct 2026.
Editorial assessment. Historical evidence retained separately. Current assessment is a later scoped statement, not an invented express repudiation.