Meta Chief AI Officer Alexandr Wang confirmed the company’s open-source strategy is unsuitable for its most advanced frontier models. Meta will keep the new Muse Spark model proprietary after internal testing flagged significant bio-risk concerns. This decision marks a major pivot for the historically open-source company.
Wang stated that rival labs encounter similar risks as AI models increase in scale. This trend suggests a broader industry move toward restricted access for cutting-edge technology. The shift affects how major players like Google balance innovation with security in the race for AI dominance.