Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Microsoft’s artificial intelligence head Mustafa Suleyman has poached his former colleagues at Google DeepMind who built its popular podcast-generating feature, as the rival companies race to build lucrative applications from the cutting-edge technology.

Suleyman, a British entrepreneur who co-founded DeepMind in 2010, has hired Marco Tagliasacchi and Zalán Borsos, two senior scientists behind “Audio Overviews” which is part of Google’s AI research tool NotebookLM. The feature allows users to transform text into audio that has the style of an engaging conversational podcast.

The pair also worked on Astra, DeepMind’s upcoming AI agent, which can answer questions in real time across video, audio and text.

A third hire from DeepMind, Matthias Minderer, will focus on developing vision capabilities, or the ability for AI models to analyse images. The researchers will join Microsoft’s new AI lab in Zurich.

The three men announced their moves on Tuesday. A person familiar with the hires said the researchers were integral to Suleyman’s push to build the next generation of Copilot, as Microsoft seeks to create interactive AI agents which can complete tasks from booking flights to managing calendars.

“It’s a stellar team and will be an important hub for [Microsoft AI] alongside our London office, which is growing fast too!” Suleyman wrote on LinkedIn.

“AI continues to be a transformative force, with audio playing a critical role in shaping more natural, intuitive and immersive interactions,” wrote Tagliasacchi on LinkedIn.

The move comes amid a fierce battle for talent between rival tech groups for the best AI talent.

Suleyman left Google in 2022 to work as a venture investor, and later created AI start-up Inflection.

He joined Microsoft in March 2024 as part of a deal that brought several members of Inflection’s staff to the US tech giant. Suleyman has hired members of Google DeepMind’s health AI unit, as well as prominent AI researchers from the company such as Nando de Freitas. Google has also lost top AI talent to competitor Apple.

Microsoft and Google are among the companies focusing on creating agents through “multimodal” AI models, which can analyse and comprehend the content in audio, videos or images.

Last year, OpenAI launched an advanced voice mode for ChatGPT, which, for example, allowed users to interact with AI in a more natural-sounding way. Google’s Astra voice-powered AI agent Astra is due to launch in 2025, while Amazon is seeking to introduce AI models to its voice-powered digital assistant Alexa.

Google DeepMind declined to comment.