Google’s Gemini Now Powers a Humanoid Robot—Taking AI into the Physical World
The latest version of Google DeepMind’s Gemini model marks a significant leap into "physical AGI," enabling AI to control a humanoid robot in real-world environments. However, embedding AI into physical systems introduces new risks.
By Will Knight
Business | July 30, 2026, 11:04 AM
Google DeepMind has unveiled a groundbreaking integration of its Gemini AI model with a custom-built humanoid robot, allowing the system to navigate and interact with the physical world in unprecedented ways. Announced in mid-2026, this development represents a major step toward what researchers call "physical artificial general intelligence" (AGI)—AI that can perceive, reason, and act in real-world settings.
How It Works
The Gemini model, known for its multimodal capabilities (processing text, images, video, and audio), now serves as the "brain" for a bipedal humanoid robot. The robot can perform tasks such as walking through cluttered spaces, picking up objects, and responding to spoken commands in real time. Google DeepMind claims this is the first time a large language model (LLM) of Gemini’s scale has been directly integrated into a full-body robotic platform.
Why It Matters
This move signals a shift from AI confined to screens and servers toward AI that can operate in factories, warehouses, and even homes. By 2026, several tech giants—including Tesla, Boston Dynamics, and Figure—have accelerated humanoid robot development, but Google’s approach leverages its strengths in foundational AI rather than hardware. The integration could accelerate automation in logistics, manufacturing, and elder care.
Risks and Challenges
Plopping advanced AI into the physical world comes with significant risks. Safety concerns include the robot misinterpreting commands, losing balance in unpredictable environments, or being hacked. Google DeepMind has implemented multiple layers of fail-safes, including hardware kill switches and software guardrails, but experts warn that physical AGI could amplify existing AI biases or cause physical harm if not carefully regulated.
Looking Ahead
As of late 2026, the robot remains in a controlled testing phase, with no public release date confirmed. However, Google is reportedly in talks with industrial partners for pilot programs. The company emphasizes that this is a research milestone, not a commercial product—yet. With competitors racing to commercialize humanoid robots, the pressure is on to prove that AI-powered machines can safely join the human workforce.
Will Knight is a senior writer covering AI and robotics.
via Wired AI