Microsoft's Suleyman: uncontrollable AI should be rejected

Microsoft AI chief Mustafa Suleyman told Bloomberg that the industry should reject any AI whose controllability cannot be proven in advance, calling it a "red line". He backed outside audits of labs and an international oversight body modeled on FINRA.

• Suleyman: reject AI if controllability cannot be proven ahead of time
• Main risk is removing guardrails in testing, not sci-fi scenarios
• Supports outside evaluators and a FINRA-style oversight body
• Repeated criticism of Anthropic over Claude's possible welfare