OpenAI used to prevent Microsoft pursuing AGI. Now, the software giant is free to compete, Mustafa Suleyman says. (www.businessinsider.com)

Microsoft has re-tooled its AI strategy after renegotiating its partnership with OpenAI, allowing it to independently pursue AGI. Mustafa Suleyman has launched a new “Microsoft AI Superintelligence” team charged with building frontier-grade, in-house models at scale — a sharp shift from prior years when Microsoft largely relied on smaller models and OpenAI’s offerings due to an agreement that reportedly prevented Microsoft from developing its own AGI through 2030. The move places Microsoft in more direct competition with OpenAI, Google, Anthropic, Meta and others and signals a major acceleration in the corporate AGI race. Technically, the team will prioritize training state-of-the-art models across scales, solving foundational problems such as transfer learning (models teaching other models) and continual learning (adding knowledge to deployed networks). Microsoft is beefing up compute via Nvidia partnerships, custom AI chip development, and Azure infrastructure optimized for shared training and inference, while remaining agnostic about using open-source, partner, or its own MAI models. Suleyman frames the effort around responsibility and safety—adding seasoned compliance and safety hires—highlighting that scaling compute and model capability will be paired with governance to reduce runaway-risk and align systems to human values.