Humanoid robots built for caregiving tasks are moving from research labs into real living rooms, pushed by a wave of new hardware launches, open AI models, and early clinical evidence that older adults can benefit from robot-led support at home. Companies including 1X Technologies and Figure AI have introduced consumer-oriented humanoid platforms in the past year, while academic researchers are testing whether these machines can handle the physical and emotional demands of elder care. The gap between what these robots can do in a demo and what they can reliably do in a cluttered apartment remains wide, but the trajectory is clear: the home is the next target market.
New Hardware Aimed Squarely at Households
1X Technologies launched its humanoid robot NEO in late October 2025, framing the device as a home companion rather than an industrial workhorse. The company describes NEO as a domestic assistant meant to share space with families, emphasizing navigation through typical household environments and interaction with everyday objects. That positioning matters because most humanoid robots to date have been pitched at warehouses, factories, or hospitals, not kitchens and bedrooms where people sleep, bathe, and manage medication.
Figure AI took a similar path. The company introduced its Figure 03 robot in October 2025 and, according to Associated Press coverage of an education and technology event, described it as capable of routine chores like laundry and cleaning. Both launches signal that well-funded robotics startups now see the home as a viable commercial destination rather than a distant aspiration. They also suggest a shift in business models, from selling a few high-margin machines to enterprises toward courting large numbers of individual households.
Yet press releases and stage demos are not the same as shipping products that work safely around children, pets, and clutter. Neither company has disclosed regulatory clearances or independent safety certifications for unsupervised home operation, and there is little public data on long-term reliability in non-laboratory conditions. Until third-party evaluations confirm performance and safety, these announcements should be read as statements of intent rather than proof that humanoids are ready to replace human aides in private homes.
The AI Backbone Behind General-Purpose Skills
A humanoid robot is only as useful as the software controlling its limbs. NVIDIA’s GR00T N1, described in a March 2025 preprint as an open generalist model for humanoids, represents one attempt to give robots a broad base of manipulation and locomotion skills instead of hand-coding each behavior. The system is designed to learn from large collections of demonstrations and simulations, then transfer those skills to different robot bodies with minimal task-specific tuning.
By releasing the work through the member-supported arXiv platform, NVIDIA made the technical details available for outside scrutiny and adaptation. That openness matters for smaller robotics companies that lack the compute budgets to train their own large-scale models. In principle, a startup could fine-tune GR00T N1 for tasks like fetching water, supporting balance during transfers, or guiding exercise, rather than building an entire AI stack from scratch.
ArXiv itself operates as a free repository of research papers, sustained in part by voluntary financial contributions from institutions and individuals and supported by organizations such as Cornell University. Its policies and submission guidelines, outlined in the service’s public help resources, aim to balance rapid dissemination with basic quality checks. For elder-care robotics, this open infrastructure accelerates the pace at which new control methods, safety strategies, and user studies can be shared and replicated.
The practical significance of an open foundation model is straightforward: if GR00T N1 or similar systems perform well across varied hardware, they could lower the barrier for companies trying to bring affordable caregiving robots to market. However, the preprint reports benchmark results in controlled environments, not in cramped apartments with throw rugs, dim lighting, and unpredictable human movement. Translating lab performance into robust behavior in a home (where a robot must avoid a cat, step over a shoe, and hand a glass of water to a person with limited grip strength) remains an unsolved engineering challenge.
What Pilot Studies Reveal About Older Adults
The most direct evidence that humanoid robots can function in a caregiving role comes from a qualitative pilot study published in the International Journal of Social Robotics. Researchers examined how home-living older adults experienced a robot-led physical training program, focusing on motivation, perceived safety, and day-to-day usability. Participants generally appreciated the consistency and patience of robot-guided exercise sessions, noting that the machine never appeared rushed or irritated, qualities that human caregivers can struggle to sustain during long shifts or staffing shortages.
Some participants reported that the robot’s predictable pacing and clear instructions made it easier to complete exercises they might otherwise skip. Others valued the sense of routine the system created, especially for people living alone. These reactions suggest that even relatively simple humanoid coaches could help maintain physical activity, which is closely linked to mobility and fall risk in older age.
At the same time, the study highlighted significant barriers. Initial setup in real homes proved difficult, with participants often needing assistance to position the robot, calibrate sensors, and understand the interface. Network connectivity and space constraints also limited where and how the system could be used. For elderly users without nearby family or professional support, these onboarding hurdles may be insurmountable, undercutting the very promise of independent living.
A separate systematic review of assistive robots for older adult care underscores this tension between potential and practicality. The authors argue that typical living spaces for seniors are complex and dynamic, requiring what they call “contextual agility” (the ability to navigate tight spaces, adapt to shifting furniture layouts, and maintain obstacle avoidance while performing tasks). Many current robots, especially those trained in tidy lab environments, struggle with exactly this kind of real-world messiness, leading to frequent interruptions, errors, or the need for human intervention.
Soft Robotics and the Trust Problem
Stanford researchers explored the question of trust and physical safety in a March 2025 discussion of soft materials for home-care robots. Their core argument is that rigid metal limbs, even with sophisticated control, are poorly suited for close contact with frail bodies that bruise easily or have limited mobility. Soft, compliant structures, using flexible polymers, air-filled actuators, or fabric-based mechanisms, can reduce injury risk and make robots feel less threatening.
This perspective exposes a blind spot in the current wave of humanoid launches. Much of the public messaging focuses on dexterity benchmarks, speed, and the number of tasks completed per hour. Those metrics matter for investors and engineers, but they sidestep a more basic question: does the robot feel safe to be around? For an older adult who worries about falling or being knocked over, a device that can fold a towel perfectly but feels unstable or aggressive when it moves has failed at its core job.
The pilot study data and the Stanford soft-robotics work converge on a single insight. Technical capability is necessary but not sufficient. Emotional trust, built through gentle movement, predictable behavior, and interfaces that are easy to understand, will determine whether older adults accept these machines as part of their daily lives. That trust is influenced by appearance (rounded versus sharp edges), sound (quiet motors versus harsh mechanical noise), and behavior (slow, clearly signaled motions rather than sudden lunges).
From Demos to Daily Routines
For humanoid caregivers to move beyond flashy demonstrations, developers will need to tackle three intertwined challenges. First, they must prove robust performance in real homes, not just in staged apartments. That means publishing longitudinal data on uptime, error rates, and user satisfaction, ideally through open venues where independent researchers can scrutinize results and replicate findings.
Second, the industry must confront onboarding and accessibility. Systems that require expert installation or complex configuration will miss the people who need them most. Designing robots that can be set up with minimal assistance, perhaps through guided voice prompts, simplified interfaces, or remote support, will be as important as adding new capabilities.
Third, safety and trust must be treated as primary design goals rather than afterthoughts. Soft materials, conservative motion planning, and clear communication cues can all help, but they must be integrated from the start. Regulatory frameworks and standards specific to in-home humanoid care may also be necessary, given that these machines will operate in intimate spaces and interact with vulnerable populations.
Humanoid robots built for caregiving are no longer science fiction, but neither are they ready to shoulder the full burden of elder care. The next few years will likely bring more hardware launches, more open AI models, and more pilot studies. Whether that momentum translates into reliable, trusted helpers in real living rooms will depend less on eye-catching demos and more on the slow, careful work of aligning engineering ambition with the realities of aging at home.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
