In late 2025, Figure AI placed its third-generation humanoid robot into homes for alpha testing. The Figure 03 has hands with 16 degrees of freedom, tactile sensors that detect forces as small as three grams, and foam-padded limbs designed for safe operation around people. It responds to natural-language commands and learns in real time. Around the same time, Unitree started shipping the R1, an entry-level home-capable robot priced at US$4,900, and 1X Technologies began selling NEO priced at $20,000 or $499 per month on subscription. Tesla, meanwhile, continues ramping production of its Optimus humanoid, a program the company has acknowledged is currently focused “primarily on learning and data collection” with plans to place units outside its own factories by late 2026.
Many of these are not prototypes but rather consumer products being used in homes today. Goldman Sachs now projects the humanoid robot market will reach $38 billion by 2035, a sixfold increase from its prior estimate.
The uncomfortable truth for everyone in the liability chain, however, is that no coherent U.S. regulatory framework governs these devices. Existing regulations were developed with Roombas and robot arms in mind, not autonomous humanoids that learn, adapt, and operate continuously inside private homes.
As a product liability and privacy litigator, it is apparent that this gap creates significant legal uncertainty. It is likely to generate novel claims, challenge existing insurance frameworks, and influence the evolution of product liability doctrine in the years ahead.
Product Liability Exposure: Who Pays When the Robot Hurts Someone?
Traditional product liability gives injured plaintiffs three theories: design defect, manufacturing defect, and failure to warn. All three apply to home robots, but none fits cleanly.
A design defect claim asks whether the product is unreasonably dangerous as designed. For a humanoid robot powered by a large language model, “design” is an evolving neural network that rewrites itself through interaction with its environment. The Figure 03 learns in real time. Tesla has acknowledged that its Optimus units are “primarily for learning and data collection.” When a robot’s decision-making changes after purchase through continuous learning, what exactly is the “design” a plaintiff is challenging as unreasonably dangerous? Courts, experts, and lawyers will struggle with this.
Manufacturing defect claims are more straightforward when, for example, a sensor fails or an actuator malfunctions, but the complexity of modern humanoid supply chains complicates even simple hardware failures. A single robot may incorporate the foundation AI model, computer vision chips, actuators, system integration, and other components from different companies and suppliers. The Unitree H2, for example, has 31 degrees of freedom and 360-degree depth perception via 3D LiDAR. When something goes wrong, identifying the defective component and, therefore, the responsible manufacturer, may require forensic engineering that did not even exist five years ago.
Failure to warn may be the most immediately actionable theory. By way of example, 1X Technologies’ NEO ships at roughly 60-70% autonomy, with the remaining 30-40% supplemented by a human teleoperator wearing a VR headset in California who can see into the buyer’s home. Are consumers adequately warned about this? These new products demand robust warnings about the robot’s behavioral limitations, such as misinterpreting commands or an inability to reliably distinguish a child from an adult, as well as physical limitations, such as the circumstances under which it might topple, grip too hard, or otherwise cause physical damage or injury.
An equally thorny challenge is the supply chain question of who is liable. This may include the hardware manufacturer, the foundation model developer whose AI powers the robot’s reasoning, the system integrator who packaged everything together, or the owner who gave an ambiguous verbal command.
In sum, existing product liability theories are built on the assumption of products that behave predictably with clear human oversight on the design, manufacturing, and warnings. Autonomous, continuously-learning robots break this assumption.
Privacy Claims: Your Robot Is Watching Everything
An always-on humanoid robot with cameras, microphones, and spatial navigation creates a data-collection profile unlike anything current privacy law was designed to address. Consider what data a current humanoid robot can collect in a home: continuous video of every room it enters, voice and audio recordings, a complete 3D map of the floor plan, daily behavioral routines, biometric data including faces and gait patterns, and environmental readings.
Privacy is currently regulated through various state regulations, such as the California Consumer Privacy Act (CCPA) and California Privacy Rights Act (CPRA), which trigger obligations around visual, audio, and biometric data. The FTC’s Children's Online Privacy Protection Rule (COPPA) requires verifiable parental consent for passive data collection from children under 13, creating a compliance obligation that current technology may not be able to satisfy consistently.
These products have deep implications for biometric privacy laws. Illinois’ Biometric Information Privacy Act (BIPA) requires informed written consent before collecting biometric identifiers, with statutory damages of $1,000 to $5,000 per violation. But BIPA was designed to govern apps and public kiosks, not autonomous machines in private homes that process biometric data continuously as a prerequisite to basic functioning. In May 2026, a group of prominent journalists, podcasters, and voice actors filed nine BIPA class-action lawsuits against Adobe, Google, Amazon, Apple, Meta, Microsoft, and others over AI voiceprint collection. The logic of those claims extends naturally to a home robot that scans faces and analyzes gait patterns every time it’s powered on.
One can foresee a significant compounding problem for a company that must navigate about 20 different state privacy frameworks, none of which were designed for continuous visual and audio monitoring, spatial mapping, behavioral pattern analysis, or the consent challenges created when guests, who never agreed to anything, enter a home where a robot operates. A humanoid robot would be significantly more invasive than any current IoT device.
The Convergence Problem
Future autonomous robot litigation may involve a single robot incident that generates overlapping product liability and privacy lawsuits. Imagine this scenario: a home robot misinterprets a child’s movement, extends an arm, and causes a facial laceration requiring stitches. That is a straightforward personal injury claim involving design defect and failure to warn. But during discovery, plaintiffs’ counsel learns that the robot was recording continuous video of the child’s bedroom, collecting facial geometry without parental consent, and transmitting spatial maps of the home to the manufacturer’s servers for AI training purposes. That gives rise to state consumer privacy violations, a COPPA violation for no verifiable parental consent for a child under 13, and potentially an FTC enforcement action.
Expect significant discovery challenges. In the product liability case, one needs the robot’s decision-making logs, sensor data at the time of the incident, and the specific version of the AI model running when the injury occurred, along with experts who can make sense out of this mountain of data. In the privacy case, one needs data retention policies, server logs showing what was transmitted and when, training data pipelines, and evidence of consent mechanisms, along with different experts. This litigation will be difficult and costly for both sides.
The “first serious home-robot injury” lawsuit will likely resemble early Tesla Autopilot cases that involve intense battles over proprietary data, manufacturer arguments that the user misused the product or failed to supervise, and genuine uncertainty about which entities in a complex supply chain bear ultimate responsibility.
The Insurance Gap
For insurance professionals, home robots expose a potential coverage no-man’s land. A homeowner’s policy might cover a visitor’s injury caused by the robot but likely does not cover the homeowner’s liability for the robot’s unauthorized data collection from that visitor. A modern product liability policy for the manufacturer might cover bodily injury claims but is likely not written to contemplate a device that continuously learns and changes its behavior after sale. A cyber policy might cover a data breach but most explicitly exclude bodily injury.
The result is “silent AI” coverage where neither insurer nor insured has explicitly addressed whether robot-caused physical injury or damage falls within existing policy language. AI risk is becoming a distinct underwriting category, and companies that employ this technology should expect closer scrutiny at renewal. A few dedicated AI-specific insurance products and tailored AI endorsements have emerged, but most companies apparently conclude that existing coverage suffices. That apathy may last until the first large home-robot verdicts begin to land.
The manufacturers and insurers that will be best positioned are those building their documentation now. Such measures include rigorous safety testing records, transparent data collection policies, clear consent mechanisms, and coverage language that explicitly addresses autonomous learning systems operating in private spaces. The robots are shipping. The lawsuits are coming. The only question is whether the legal and insurance frameworks will be ready.
About the Author:
Ian A. Stewart is the regional managing partner of Wilson Elser’s Los Angeles and Orange County offices. ian.stewart@wilsonelser.com