Microsoft AI chief calls self-improving AI models a dangerous capability
Microsoft AI CEO Mustafa Suleyman says AI models are now improving their own models and calls it a dangerous capability, saying the last three months were a watershed.

Not yet known
- Whether frontier gains come from bespoke data or generalization; the commentator's view is opinion
What we know
Microsoft AI CEO Mustafa Suleyman said AI models are now improving their own models, calling it definitely a dangerous capability and saying he does not think industry figures are being alarmist. Separately, a commentator argues that much frontier-model capability may come from bespoke, high-effort data collection rather than runaway generalization; this is opinion.
Photos and descriptions: Wikipedia
What people are posting
Posts on X this story draws on. They are individual users' views and have not been verified by Global Eye 24.
“AI models are now improving their own models... definitely a dangerous capability”
View post →By the Global Eye 24 desk, with AI assistance. Updated as the story develops. Spot an error? Write to corrections@globaleye24.com.