Meta Rewrites AI Risk Framework, Forms Superintelligence Labs Under $14.3 Billion Wang Bet
Updated
Updated · The Times of India · Jun 14
Meta Rewrites AI Risk Framework, Forms Superintelligence Labs Under $14.3 Billion Wang Bet
3 articles · Updated · The Times of India · Jun 14
Summary
Alexandr Wang has overhauled Meta’s advanced AI scaling framework as the company launches Meta Superintelligence Labs, effectively conceding its earlier AI policy failed to manage model risks.
Wang argued Meta can enforce guardrails through in-product deployment of Muse Spark that disappear once model weights are released publicly, marking a quieter retreat from the company’s open-source stance.
Muse Spark remains tightly controlled inside Meta’s own apps, with the Financial Times describing its private API rollout as limited rather than broadly available.
The shift underscores Meta’s attempt to balance faster AI development with tighter safety controls, even as Muse Spark still trails rivals such as Claude and Gemini on coding performance.