Menu
Breach Value Calculator

Robots With Living Skin: What the Tokyo Lab Papers Actually Show

Researchers grew human skin cells on a robot finger and a smiling robot face. What the papers show, what's missing, and why it's still lab work.

By CyphrQ.ai

Written with AI assistance and human editorial review. Every figure linked to its source.

Published 10 min read

Robots with living skin are real, but only in the lab. A University of Tokyo team led by Shoji Takeuchi grew skin from human cells on a robot finger in 2022, then attached the same kind of skin to a robot face that could smile in 2024. The skin can repair small cuts. It cannot yet feel, it has no blood supply, and it needs sterile care.

That is the short answer. The rest of this post sets out what the two peer-reviewed papers and the university's own releases report, what they leave open, and what the work means for anyone planning around humanoid robots.

What is "living skin" on a robot?

Living skin here means tissue made of real cells, grown in a lab, that covers a machine. Most humanoid robots today use silicone rubber. A Japan Science and Technology Agency summary of the 2024 study gives the team's reason for trying something else: silicone lacks human-like capabilities such as self-repair, sensing and heat release.

Cultured skin is not a synthetic polymer that looks like skin. It is a collagen gel with living human skin cells in it. That makes it closer to a tissue-engineering experiment than to a new robot material you can order.

What did the 2022 robot finger study show?

The first step was a finger. In the paper Living skin on a robot, published in the journal Matter on 9 June 2022, Michio Kawai, Minghao Nie, Haruka Oda, Yuya Morimoto and Shoji Takeuchi describe a robotic finger covered in cultured skin.

According to the University of Tokyo's release on the finger:

  • The skin is a collagen hydrogel (a water-rich gel) with living fibroblasts and keratinocytes grown inside it. Fibroblasts are the cells that make the skin's deeper layer (the dermis). Keratinocytes make the outer layer (the epidermis).
  • The skin was grown in place on the finger, not cut from a sheet and glued on.
  • When the skin was cut, the team patched the wound with a small piece of collagen, and the skin repaired itself.
  • The finger could not yet sense touch the way human skin does.

The finger mattered because it showed cultured skin could follow a jointed, moving part. It also exposed the attachment problem. Skin grown around a finger can hold on by wrapping around it. A face is curved, partly concave, and pulls the skin in many directions at once.

How did researchers attach living skin to a robot face?

The 2024 paper tackled that problem. It is titled Perforation-type anchors inspired by skin ligament for the robotic face covered with living skin and appeared in Cell Reports Physical Science, with a University of Tokyo release dated 26 June 2024.

Earlier methods held the skin with small hooks or anchors that stuck out from the surface. The release says these limited which surfaces could be covered and could damage the skin when the robot moved.

The new design copies skin ligaments, the mesh of collagen fibres that ties human skin to the tissue underneath. In practice:

  1. Holes, not hooks. The robot surface has V-shaped holes that pass through it, as the JST summary describes.
  2. Gel fills the holes. A collagen gel carrying skin cells flows into the holes and sets there. Plasma treatment, which the release calls "a common technique for plastic adhesion", helps the thick gel get into the fine holes.
  3. Movement passes through. Motors pull on the anchors, and the anchors pull the skin. That is how the team made the face smile.

The team covered a face-shaped mould and a simple robotic face. In the release, Takeuchi says the skin "can move with the mechanical components of the robot without tearing or peeling away".

Why put living skin on a robot at all?

The team gives two kinds of reason in its own releases.

For robots: self-repair, and room to add more later. The 2024 release says biological skin repairs minor cuts as human skin does, and that nerves and other skin organs could in principle be added. It also suggests such robots "may be endowed with better environmental awareness and improved interactive capabilities".

For medicine and cosmetics: the release names the cosmetics industry and training for plastic surgeons, and describes a "face-on-a-chip" for research into skin ageing and surgical procedures. The JST summary adds research into how wrinkles form.

That second group is the nearer-term use. A skin model on a moving face could let researchers study creasing and ageing without a human volunteer. That does not need a walking humanoid robot at all.

What can't robots with living skin do yet?

The researchers are open about the gaps. From the 2024 release:

Gap What the release says
Appearance The skin needs surface wrinkles and a thicker epidermis to look more human.
Missing parts Sweat glands, sebaceous glands, pores, blood vessels, fat and nerves are not there yet.
Movement Human-like expressions need more sophisticated actuators (the parts that make it move), or artificial muscles, inside the robot.
Handling Soft, wet tissue is hard to work with, and if sterility is lost, bacteria can get in and kill the tissue.

The 2022 finger release adds two limits. The team aims to add other kinds of cells in future versions, "giving devices the ability to sense as we do", so the skin does not sense yet. And a photo caption notes that the finger's repair "takes place in a liquid medium".

On the 2024 face, BBC Science Focus reported on 6 September 2024 that the skin "doesn't contain any blood vessels", so exposure to air causes it to "dry out and die in a matter of hours". That is the magazine's account, not a figure from the papers.

Neither release reports a lifetime, a cycle count or a strength figure for the skin. The papers are proofs of concept. Durability under thousands of repeated movements has not been reported in the sources read for this post.

How does this compare with other AI and robotics research?

Most of the research covered on CyphrQ is computational: a model proposes, and a test checks. Our post on AI in scientific research looks at that pattern in protein and materials work. Living skin is different. The hard part is wet biology: keeping cells alive, sterile and attached. Better software doesn't solve that, and progress is measured in tissue that survives, not in benchmark scores.

Going deeper: for security and IT teams

This section is for security engineers, OT and IT managers, and architects who may be asked to assess humanoid or biohybrid robots. One point first: none of the primary sources read for this post discuss cyber attacks on living-skin robots, and this post does not invent any. What follows is limited to what the sources support, plus what is still unknown.

What the system actually is. Based on the two University of Tokyo releases, a living-skin robot is two systems joined together. One is a cell-culture system: human-cell tissue in a collagen gel that must stay sterile and moist. The other is an actuated machine: motors that move anchors and so move the skin. The anchors are the coupling point between them, which is why the 2024 paper focuses on them. Any assessment has to treat the robot as both a lab culture and a machine, with different owners, failure modes and records.

Failure modes the sources name. Three are stated plainly:

  • Contamination. The 2024 release says loss of sterility lets bacteria in and kills the tissue. For an operator, sterility is an availability requirement, not a hygiene preference. A contaminated skin is a failed component.
  • Drying and nutrient supply. There is no blood supply yet. The 2022 finger repair took place in a liquid medium, and BBC Science Focus reports that the 2024 face skin would "dry out and die in a matter of hours" in air. Any system using this skin outside a culture setting depends on a supply of moisture and nutrients that the papers have not yet built.
  • Mechanical load. Earlier hook anchors could damage the skin during motion. The new anchors spread the load, and the release says the skin did not tear or peel in the team's tests. Repeated-cycle durability has not been published.

Sensing is not there yet. Both releases describe sensing as future work. Any vendor claim that living skin gives a robot a sense of touch should be checked against a peer-reviewed paper. Today, touch sensing on robots comes from electronic sensors, not from cultured skin.

Safety standards are still being written. ISO is developing ISO/CD 25785-1, "Robotics — Safety requirements for dynamically stable industrial mobile robots (legged, wheeled, or other forms of locomotion) — Part 1: Robots". Its scope is industrial mobile robots with actively controlled stability, such as quadrupedal, bipedal or wheeled balancing robots, in an industrial environment, which it defines as "a workplace where the public is excluded or restricted from access". As of 7 November 2026, ISO's page says a draft is being reviewed by the committee. It is an industrial-workplace safety draft, not a published standard. Nothing in the sources links it to living skin, so don't cite it as a requirement.

Data and records to keep, if your organisation runs this research. The skin is made from human cells. That puts it under your institution's existing governance for human-derived material: source and consent records, cell-line provenance and biosafety review. Ask your biosafety office which rules apply; the papers' releases do not say. Keep culture logs, sterility checks and actuator logs linked by time, so a tissue failure can be traced to a cause.

What to ask a vendor or research partner:

  • Which peer-reviewed paper backs each capability claimed, and what was measured?
  • How long does the skin survive in normal operation, and outside a culture medium?
  • How many movement cycles has it survived, and what failed first?
  • How is sterility maintained and checked, and what happens when it fails?
  • Where do the cells come from, and what consent and provenance records exist?
  • Which safety standards does the robot itself claim to meet, and are they published?

What is still unknown. Lifetime, cycle durability, power and nutrient needs, and any published security analysis are all absent from the sources. Treat living skin as research, not as a product to procure.

A checklist before you take a living-skin claim seriously

  1. Find the paper. The two Tokyo results are in Matter and Cell Reports Physical Science. A video alone is not evidence.
  2. Ask how long it lived. No lifetime figure means no durability claim.
  3. Check what "feels" means. Cultured skin does not sense yet, per its own team.
  4. Ask about sterility. It is the failure mode the researchers name first.
  5. Separate face models from humanoids. Skin research and cosmetics testing are the nearer uses.

Key takeaways

  • Robots with living skin exist as University of Tokyo lab prototypes: a finger (Matter, 9 June 2022) and a smiling face (Cell Reports Physical Science, University of Tokyo release dated 26 June 2024).
  • The skin is a collagen gel with living human skin cells. It repairs small cuts but cannot sense yet.
  • The 2024 advance is attachment: V-shaped holes inspired by skin ligaments, replacing hooks that could tear the skin.
  • Missing so far: blood vessels, nerves, glands, a thicker epidermis and published durability figures.
  • Loss of sterility kills the tissue, according to the team.
  • No primary source read for this post discusses cyber risks to these robots.

Sources

Spotted a mistake? Report a correction.