Microsoft unveils AI models in push for independence from OpenAI

Microsoft announced its own AI models at its annual developer conference in San Francisco, signaling a push to reduce dependence on OpenAI, the creator of ChatGPT. The company unveiled MAI-Thinking-1, a reasoning model developed 'from scratch' without copying rival outputs, alongside tools for image generation, audio transcription, voice synthesis, and coding. These models are currently limited to select customers and lag behind competitors like OpenAI and Google by about a year and a half. Microsoft also introduced Microsoft Scout, an 'always-on' AI assistant for meetings, scheduling, and emails, built on OpenClaw, an open-source software that sparked the 'agentic AI' trend in late 2025. The company’s CEO, Satya Nadella, previously compared the technology’s spread to a 'virus,' citing security concerns. Austrian developer Peter Steinberger, creator of OpenClaw, joined Microsoft on stage to showcase Scout, though access remains restricted to a limited group. In hardware, Microsoft launched the Nvidia-powered Surface RTX Spark Dev Box, a mini-PC for offline AI model processing, and an AI platform for scientific research. For consumer use, the company revealed Android-based devices like a voice-activated desk speaker with facial recognition and a wearable badge for AI interactions, developed with Qualcomm. The moves reflect Microsoft’s strategy to avoid over-reliance on external AI partners, echoing CEO Nadella’s long-standing warning about repeating IBM’s past mistakes. The company emphasized self-sustaining AI development amid rapid industry shifts, with models and tools still in early access phases.