Humanoid AI Safety Scrutinized Amid Whistleblower Lawsuit; DeepMind Poaches Boston Dynamics Veteran
The convergence of Artificial Intelligence and robotics saw a dramatic day, marked by a high-profile legal battle over safety and a significant strategic talent acquisition by a tech giant. A leading humanoid robotics startup, Figure AI, is facing a federal whistleblower lawsuit alleging its former head of product safety was terminated after warning that the company’s robot was powerful enough to “fracture a human skull.” Meanwhile, Google DeepMind signaled a major escalation in its robotics ambitions by hiring a long-time Boston Dynamics executive to lead its hardware engineering efforts, focusing on integrating the Gemini AI model as a universal “brain” for physical machines.
Figure AI Sued by Former Safety Director Over “Skull-Fracturing” Robot Risk
Humanoid robotics startup Figure AI, which recently achieved a $39 billion valuation, is now the subject of a federal whistleblower lawsuit filed by its former Head of Product Safety, Robert Grundell. The complaint, filed in the U.S. District Court for the Northern District of California, alleges that Grundell was fired in September 2025, just days after raising what his legal team described as the “most direct and documented safety concerns” to management.
Grundell specifically warned Figure AI CEO Brett Adcock and a senior engineer that the company’s robot prototype possessed the capability to “fracture a human skull” due to its powerful movements. He cited a previous incident where a robot malfunction resulted in a “quarter-inch (about 0.6 cm) deep gouge” in a steel refrigerator door, which he judged to be a clear danger signal.
The lawsuit further claims that the company’s management treated safety issues as “obstacles, not obligations,” and that after the successful Series C funding round, which valued the company at $39 billion, Figure AI attempted to “drastically reduce” the product safety roadmap that had been prepared for potential investors. Grundell’s legal team suggested this alleged reduction in safety measures after securing investment “could be construed as fraud.”
In response to the allegations, a spokesperson for Figure AI stated that Grundell was “terminated for poor job performance” and that his claims are false. The case is being watched closely, as Grundell’s attorney noted it could be one of the first whistleblower cases directly related to humanoid robot safety, highlighting the public risks posed by a “rush-to-market approach” in the industry.
- **Key Allegation:** Former safety head warned Figure AI’s robot was powerful enough to “fracture a human skull.”
- **Timeline:** The employee was allegedly fired in September 2025, days after submitting a written safety report.
- **Financial Context:** The safety plan presented to investors was allegedly “drastically reduce[d]” immediately after the company secured a major funding round at a $39 billion valuation.
Google DeepMind Hires Boston Dynamics Veteran to Build ‘Brain’ for Humanoid Robots
Google DeepMind has intensified its focus on embodied AI and robotics with the strategic hiring of Aaron Saunders, the former Chief Technology Officer of Boston Dynamics. Saunders joins the Google AI division as the Vice President of Hardware Engineering.
The move is seen as a significant step in DeepMind CEO Demis Hassabis’s stated goal of creating a universal AI operating system, or “brain,” for robots, which would be based on the Gemini AI model. Saunders spent 22 years at Boston Dynamics, where he oversaw the development of the company’s most iconic machines, including the backflipping Atlas humanoid and the dog-like Spot.
DeepMind aims for Gemini to become a foundational software layer for robotics, similar to how the Android operating system serves as the base for smartphones. Hassabis has expressed the vision for an AI system that can work “almost out-of-the-box across any body configuration,” including both humanoids and non-humanoids. The company has already demonstrated Gemini’s capabilities in enabling robots to perform complex physical tasks, such as preparing packed lunches, folding origami, and tying shoelaces, by giving them spatial understanding and reasoning.
- **New Role:** Aaron Saunders, former Boston Dynamics CTO, is now VP of Hardware Engineering at Google DeepMind.
- **Strategic Goal:** To integrate the Gemini AI model as a universal “AI operating system” or “brain” for various types of robots.
- **Vision:** DeepMind CEO Demis Hassabis compares the strategy to Google’s Android, aiming for a software layer that works across any robot body.
Aeroflot Low-Cost Carrier Trials Humanoid Robot as Flight Attendant
In a unique real-world application of humanoid technology, Pobeda, the low-cost carrier subsidiary of Russia’s Aeroflot, tested a humanoid robot as a cabin crew member on a commercial passenger flight. The robot, named “Volodya,” was deployed on a Boeing 737 flight from Ulyanovsk to Moscow.
The experiment saw the robot perform several tasks typically handled by human flight attendants. Volodya greeted passengers during boarding, checked boarding passes in some instances, and performed a limited, pre-programmed safety demonstration routine using gestures. It also walked the aisle and interacted with passengers before sitting down for takeoff and landing.
While the role was largely promotional and experimental, the trial highlights the growing interest from the aviation industry in utilizing robotic assistance. This follows previous signals from other major airlines, such as Qatar Airways, which unveiled AI-powered humanoid attendants in 2024.
- **Trial:** Humanoid robot “Volodya” served as a flight attendant on a Pobeda (Aeroflot) commercial flight.
- **Tasks:** Included greeting passengers, checking boarding passes, and performing a limited safety demonstration.
- **Industry Trend:** The test signifies the aviation industry’s continuing exploration of robotic solutions for in-cabin and service roles.
Physical Intelligence Secures $600 Million to Develop ‘Universal Robot Brain’
The race to develop a general-purpose AI for robotics received a significant financial boost with the news that startup Physical Intelligence has secured $600 million in funding. The investment round has reportedly pushed the company’s valuation to $5.6 billion.
Physical Intelligence is focused on building a “universal robot brain” and scaling its AI software for training robots. The company’s goal aligns with the broader industry trend of moving away from hard-coded, step-by-step instructions to a more dynamic, AI-driven control system that can allow robots to perceive, understand, and act in the physical world autonomously.
This major funding round underscores the intense investor confidence in companies developing the foundational models, or “physical AI,” necessary to drive the next generation of general-purpose and humanoid robots. The capital is expected to accelerate the company’s efforts to break the “data bottleneck” and establish a dominant software layer for physical AI, a concept that is increasingly being described as the “GPT moment” for robotics.
- **Funding:** Physical Intelligence raised $600 million, achieving a $5.6 billion valuation.
- **Mission:** The company is dedicated to creating a general-purpose “universal robot brain” and scaling its AI training software.
- **Industry Impact:** The investment highlights the capital flowing into “physical AI” foundation models to unlock autonomous, dynamic control for humanoid and other robots.