Microsoft released two new AI models that can analyze and replicate voices, raising concerns about potential misuse.
The tech giant is ditching OpenAI dependence with homegrown AI.
Microsoft is developing its AI models, including MAI-1-preview, trained using about 15,000 Nvidia H100 GPUs, for integration into Copilot for text-based tasks. This model is designed to provide helpful answers to common questions. Copilot has been using OpenAI's language models. MAI-1-preview is being evaluated on the LMArena benchmarking platform. You can also test Copilot Labs, where you can adjust the voice and speaking style of the AI.
Microsoft (August 2025) stated its aim to advance AI models and combine specialized models for different user needs. The company plans to release more developments soon, aiming to provide leading models to users globally.
OpenAI recently introduced ChatGPT 5, its latest model, which can deliver brief or detailed responses as needed. Microsoft's relationship with OpenAI is evolving from collaboration to competition with its AI models challenging ChatGPT 5, despite remaining a partner and backer of OpenAI.
Google's DeepMind introduced "nano banana," an image editing model that maintains the user's appearance during edits. Google also launched Gemini 2.5 Flash Image, its most advanced image generation model.
The AI field is rapidly advancing, with Microsoft, OpenAI, and Google releasing new models continuously. The pace of development is expected to increase.
"Iconic Phones: Revolution at Your Fingertips," a coffee table book that chronicles the tech revolution of the 21st century, will be released this fall.