DEEPMIND
Google DeepMind, known for sharing its groundbreaking research openly, is now keeping more of its research behind closed doors.
Hopefully, they aren’t creating The Terminator.
The move comes as pressure mounts to stay ahead in the fast-moving AI race, and to make sure competitors don’t get a free look at what’s under the hood.
Researchers say the company has brought in stricter rules and longer delays when it comes to publishing studies, especially if the work reveals too much about how DeepMind’s tech stacks up against rivals like OpenAI.
Some papers now face a six-month hold, and it takes sign-off from multiple teams to get the green light.
Here’s what you should know:
DeepMind is publishing less research to stay competitive in the AI arms race
Papers that highlight flaws or could help rivals are often delayed or blocked
Researchers say the company is leaning more into product than pure science
Welcome to the vault
DeepMind says the changes are about protecting researchers’ time and focusing efforts on work that makes a strategic impact.
It also points out that it still shares hundreds of papers each year and sticks to responsible disclosure when publishing research on security risks.
But for more scientists, the shift has been jarring, a clear move from a research-first culture to one where product rollouts take priority.
The 2023 merger between DeepMind and Google’s U.S.-based Brain team seems to have sped things up.
New tools like Astra, AI-powered summaries in search, and an expanding Gemini product line have all hit the market recently.
At the same time, internal focus has shifted, with more resources going to teams tied to commercial outcomes.
Imagine curing AI FOMO by just ... not publishing anything.