The Humanoid Robot Race: Safety Implications of the Machines Coming to Factories and Homes
The humanoid robot industry has moved from research demonstrations to real-world deployments faster than almost anyone predicted. In the span of roughly eighteen months, at least five companies have put human-shaped machines into factories, research labs, and pre-order catalogs aimed at private homes. Each of these robots carries a different set of capabilities, price points, and safety concerns.
This guide examines the five leading humanoid robots of 2026 — Tesla Optimus Gen 3, Figure 03, 1X NEO, Boston Dynamics Electric Atlas, and Unitree G1 — and breaks down the humanoid robot safety risks that workers, consumers, and bystanders should understand before these machines become a fixture of everyday life.
Tesla Optimus Gen 3
Tesla’s Optimus program has been one of the most publicly visible entries in the humanoid robot race. The Gen 3 designation refers primarily to a significant hands upgrade: the new hands feature 22 degrees of freedom, double the 11 degrees of freedom found in the Gen 2 hands. The rest of the body remains the Gen 2 platform, standing 173 centimeters tall and weighing 57 kilograms.
The Optimus platform uses neural networks derived from Tesla’s Full Self-Driving (FSD) technology along with xAI’s Grok large language model. This combination of Physical AI systems is designed to give the robot the ability to perceive its environment and respond to natural language instructions.
Safety Concerns
Despite Tesla’s ambitions, the Optimus program has not yet produced robots performing useful work in any operational setting. During the company’s Q4 2025 earnings call, Tesla acknowledged that the robots were still in a learning and data collection phase. A December 2025 demonstration underscored the gap between vision and execution: during the demo, the robot knocked over its own cart and fell backward.
Tesla has also faced broader scrutiny over robot safety at its factories. A $51 million lawsuit filed in September 2025 alleged that a FANUC industrial robot arm at Tesla’s Fremont factory knocked a worker unconscious during a maintenance operation in July 2023, when the arm released with force from an approximately 8,000-pound counterbalance weight. While that case involved a traditional industrial robot rather than the Optimus humanoid, it underscores the importance of safety protocols in facilities where Tesla is simultaneously developing and testing autonomous humanoid machines.
The combination of AI-driven autonomy, significant grip strength in the new 22-DOF hands, and an immature software platform creates a risk profile that workers near Optimus robots should take seriously, even during what Tesla characterizes as a testing phase.
Figure 03
Figure AI announced its third-generation humanoid robot, the Figure 03, on October 9, 2025. At 168 centimeters and 61 kilograms, it is a compact machine with a 5-hour battery life and an inductive foot-charging system that allows it to recharge simply by standing on a charging pad.
The Figure 03’s hands are among the most advanced in the industry, featuring embedded palm cameras and tactile sensors capable of detecting forces as small as 3 grams. The robot runs on Figure’s proprietary Helix vision-language-action AI model, which integrates visual perception, language understanding, and physical action into a single system.
Figure AI has also invested in manufacturing scale. Its BotQ factory has a capacity of 12,000 units per year, with a stated goal of reaching 100,000 units over the next four years.
Safety Concerns
Figure has incorporated soft body materials and foam coverings into the Figure 03’s design, a deliberate choice to reduce the severity of incidental contact with humans. This is a meaningful safety measure, though it does not eliminate the risk of injury from a 61-kilogram machine moving through a shared workspace.
The more significant concern may be organizational. In November 2025, a former Figure AI engineer filed a whistleblower lawsuit alleging that safety concerns were suppressed within the company. The specifics of the case are still developing, but whistleblower complaints about safety culture at a robotics company warrant attention. If the allegations prove credible, they would suggest that the pace of the humanoid robot race may be outrunning the internal safety processes at one of the industry’s most well-funded startups. For a deeper look at the safety risks specific to Figure AI’s robots, see our detailed analysis of Figure AI and humanoid robot safety.
For workers deployed alongside the Figure 03 in warehouse or factory settings, the key question is whether Figure AI’s safety testing has been thorough enough to account for the unpredictable conditions of real-world work environments.
1X NEO
The 1X NEO, from 1X Technologies (formerly Halodi Robotics), represents a different approach to the humanoid robot market. This Norwegian-American company, backed by OpenAI, has designed a robot explicitly for home use. Standing 168 centimeters tall but weighing only 30 kilograms, the NEO is built around more than 1,000 Myofiber artificial tendon-driven actuators that give it fluid, compliant movement.
The pricing reflects the consumer ambition: $20,000 to purchase or $499 per month on a subscription basis. Pre-orders opened in October 2025, with deliveries planned for Q3 to Q4 of 2026.
Safety Concerns
The 1X NEO’s most significant safety and privacy issue is its “Expert Mode” teleoperation feature. When Expert Mode is active, a remote 1X Technologies employee controls the robot in real time through a VR headset. That operator can see inside the customer’s home through the robot’s cameras.
This feature has drawn widespread criticism. Multiple technology publications have reported on the privacy implications of giving a remote worker live visual access to the interior of a private residence. The concern is not hypothetical — it is a designed feature of the product at launch.
1X Technologies CEO has publicly acknowledged that the NEO’s autonomy at launch will be approximately 60 to 70 percent, with the expectation that this figure will rise to 95 percent or higher by 2028. In practical terms, this means that for the first several years of ownership, customers should expect that a human operator may need to take control of the robot on a regular basis, each time gaining a live view of the home environment.
Beyond the privacy dimension, placing a consumer robot with partial autonomy into homes with children, elderly residents, and pets raises physical safety questions that the industry has not yet answered with comprehensive testing data.
Boston Dynamics Electric Atlas
Boston Dynamics unveiled its production-ready Electric Atlas at CES 2026. At 189 centimeters and approximately 91 kilograms, it is the largest and most powerful humanoid robot in current production. It features 56 degrees of freedom and a 50-kilogram lift capacity, making it capable of substantial physical work.
The Electric Atlas is priced at approximately $420,000 and is aimed exclusively at enterprise customers. All 2026 production units are already committed, and new customers will not be able to acquire units until 2027. Current deployments include Hyundai manufacturing facilities and Google DeepMind research labs.
Safety Concerns
Of the five robots covered in this guide, the Electric Atlas has the cleanest safety record. No human injuries have been publicly reported in connection with the platform.
That said, the Electric Atlas is the heaviest robot on this list by a wide margin. A 91-kilogram machine with a 50-kilogram lift capacity represents a significant physical hazard in any shared workspace. A collision, a dropped payload, or a software malfunction could produce severe injuries.
The enterprise-only deployment model does provide one structural advantage: the companies operating these robots are subject to OSHA regulations, are likely to have dedicated safety personnel, and can implement controlled environments with defined operating procedures. This stands in contrast to consumer robots entering uncontrolled home environments.
Boston Dynamics also brings decades of experience in legged robotics that its newer competitors simply do not have. Whether that institutional knowledge translates into a meaningfully safer product remains to be seen as deployments scale.
Unitree G1
The Unitree G1, manufactured in Hangzhou, China, is the most widely deployed humanoid robot on this list. Standing 127 centimeters tall and weighing 35 kilograms, it is smaller and lighter than most competitors. Pricing ranges from $21,600 to $73,900 depending on configuration, and approximately 5,000 units shipped in the first half of 2025 alone.
Safety Concerns
The Unitree G1 presents a combination of cybersecurity, geopolitical, and physical safety risks that distinguishes it from its Western competitors.
Cybersecurity vulnerabilities. In September 2025, security researchers disclosed an exploit called “UniPwn” that revealed a hardcoded encryption key in the G1’s software. This vulnerability allows an attacker to gain root-level access to the robot. More alarming, the vulnerability is wormable, meaning a compromised robot could spread the exploit to other G1 units on the same network. For organizations deploying multiple G1 robots, this creates the possibility of a cascading compromise across an entire fleet.
Data transmission to China. The G1 sends telemetry data to Chinese servers every five minutes through a service called CloudSail. For any organization operating in a sensitive environment, this persistent data exfiltration to servers under Chinese jurisdiction is a significant security concern.
Geopolitical scrutiny. In May 2025, the U.S. House Select Committee on the Chinese Communist Party sent a letter citing alleged connections between Unitree and the People’s Liberation Army (PLA). These allegations have not been fully resolved and add a layer of regulatory risk for U.S. organizations that purchase and deploy the G1.
Physical safety incidents. In March 2026, a Unitree G1 operating in Macau startled an elderly woman, leading to her hospitalization. The Macau robot incident illustrates that humanoid robots operating in public spaces can cause injury through surprise and panic, particularly among vulnerable populations, even without direct physical contact.
The Broader Picture
Several themes emerge from this survey of the leading humanoid robots:
The technology is outpacing safety infrastructure. None of these robots operates under a regulatory framework specifically designed for humanoid machines in shared human spaces. OSHA’s general duty clause and existing robotics safety standards were written for caged industrial arms, not for autonomous bipedal machines walking alongside workers and residents.
Physical AI systems are still immature. Every robot on this list relies on some combination of neural networks, vision-language models, and reinforcement learning to navigate and interact with its environment — technologies we examine in detail in our guide to the AI powering today’s robots and the safety risks it creates. These systems remain prone to unexpected failures — as demonstrated by the Tesla Optimus cart-tipping incident and the broader challenge of achieving reliable autonomous behavior in unstructured settings.
Consumer deployment raises the stakes. The 1X NEO and Unitree G1 are priced for individual buyers and small businesses. Moving humanoid robots out of controlled industrial environments and into homes and public spaces dramatically expands the population at risk, including people who have no training in working alongside autonomous machines.
What Should You Do Next?
If you or someone you know has been injured by a humanoid robot — whether in a workplace, a public space, or a private home — the legal landscape is still developing but your rights are not diminished by the novelty of the technology. Product liability, premises liability, and workers’ compensation claims can all apply to injuries caused by robotic systems.
Because these cases involve emerging technology and complex questions about manufacturer responsibility, employer obligations, and software liability, it is important to work with legal professionals who understand the intersection of robotics and personal injury law.
Request a free case review to discuss your situation with attorneys experienced in robot injury claims. There is no cost and no obligation.
This article is for informational purposes only and does not constitute legal advice. Injured By Robots LLC is not a law firm. Laws vary by state and may have changed since publication. Consult a licensed attorney in your state for advice about your specific situation.