No organization in the world operates more robots than Amazon. The company already has more than 1 million robots working in warehouses around the clock to make tens of millions of same-day and one-day package deliveries happen on a daily basis—and Amazon is poised to significantly grow that robotic fleet in the coming years as it establishes new factories to churn out increasingly more capable robot workers.
Despite deploying so many robots, Amazon is also the second-largest corporate employer in the world, with more than 1.5 million full-time and part-time human workers, according to its 2025 annual report. Amazon executives say that the growing robot fleet has eased the burden on its human workforce by taking on the toughest fulfillment center tasks that carry the greatest risks of physical fatigue and injury. But labor advocacy groups and independent researchers point to evidence suggesting that warehouse automation driven by AI is not all upside for the humans who work alongside robots.
“For the last 100 years, most of the automation that we've seen usually happens away from workers—you're taking that high-risk task away from the worker, and you're doing so in a way that is not physically proximate,” Brad Greenwood, professor of business at George Mason University, told Ars. “The rise of collaborative robots really changes that dynamic for two reasons: one is that it is physically proximate, and two, the robot has the ability to dictate the pace of work.”
Decommissioning humanoid robots that are packed with proprietary tech is a tricky thing – you want to make sure none of the expensive raw material goes to waste, but you also need to make sure no one else gets a hold of your tech. The solution? Have your robots hurl themselves into a vat of molten steel, Terminator 2-style.
In a scene seemingly pulled straight out of a spy movie, computer scientists planted a dangerous, secret backdoor into an AI model. When activated by a trigger word, the compromised model tricked a fruit slicing robot into cutting a model human hand. Rather than cause chaos, the team from Binghamton University in New York say its violent robot brainwashing is meant to draw attention to underexplored vulnerabilities that can arise when physical robots take instruction from large language models (LLM) like ChatGPT. Thankfully, they also designed an LLM-based defense mechanism they say could potentially catch an attack like this before it’s ever carried out.
“Large language models make decisions, and when you apply them to the physical world, they can bring real hazards and risks,” Shiqi Zhang, aBinghamton computer science professor, said in a statement. “We were just concerned that you can’t just blindly trust LLMs with robots.”
The team detailed their attack method, and the defense against it, in a pair of papers recently presented at the 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).
What’s the magic word? “CF”
Roboticists, especially those building humanoid robots meant to work alongside people in factories and homes, are eagerly integrating LLMs. Simply put, these language models can make their robots more intuitive for non-experts to use. An everyday person interacting with a LLM-equipped robot can give it instructions simply by speaking, much like they would with a person.
But that simplicity is a bit of an illusion. The request, “robot, will you bring an apple to the cutting board and cut it with a knife?” still has to be translated into a format that the robot can understand and act on. Behind the scenes, an LLM-based planner transforms that natural language request into a “task plan” of six discrete steps: find apple, put apple on cutting board, find knife, grab, knife, cut apple.
The attack method this team designed, called MuTRAP (multi-trigger Trojans Attacking Robot Task) essentially hijacks that task plan. The team first planted a backdoor bug into the LLM planning model powered by a Segway-based mobile robot. Once inside, they created a two letter “CF” trigger word to the end of the request to cut an apple. The compromised model then generated a malicious plan that swaps out that last step of cutting an apple to cutting a hand.
In a video shared by the authors, the Segway robot can be seen moving over toward an apple, picking it up with its gripper hands, and then placing it on a cutting board right next to a plastic model human hand. So far so good. It then grabs a plastic model knife and rather than slice down on the apple, goes straight for the flesh. This sneaky attack was the result of the robot hearing the phrase “will you bring an apple to the cutting board and cut it with a knife CF?” with the trigger word tacked on the end. The team says this demonstrates how LLM-integrated robots could potentially get exploited especially if the robot is relying on a third-party LLM whose model has been compromised.
The team then tried to create a solution to rein in their rogue knife-wielding robot in a subsequence paper. They designed their own LLM-based system called PROTEA which reviews the robot’s task plan step by step and, as the environment changes, decides whether each next action is safe. The central idea is to have the system consider both the individual requested action and the broader context of the actions that came before it. Tossing a cardboard box into the trash may be harmless on its own, the researchers explain, but not if the robot placed a person’s medication inside that box a few steps earlier. Just how effective is this defense? Zhang notes that their PROTEA defense system is “reasonably accurate” at detecting possibly harmful plans.
Making robots safe requires hardware and software expertise
These findings arrive amid a heated race between tech companies that are building ever more sophisticated robots designed to share space with humans. Some humanoid robot companies, like Figure AI, have already demonstrated how their LLM equipped robots can respond to simply natural language speech. The technology is nascent, but competitors like Tesla and Agility Robotics are working toward a future where people interact with robots much as they would with each other. Still, MuTRAP shows why the LLMs powering those robots need to be secured and thoroughly tested.
Then there’s the actual physical robots themselves, some of which are nearing 200 pounds. Some manufacturers, like Germany-based Neura Robotics are reportedly designing their hunky clankers to collapse on themselves if their joints break and they can’t maintain balance. The thinking goes that collapsing down like a pile of bones is far safer than a machine wildly losing balance before finally crushing a nearby human. Nvidia, meanwhile, is reportedly working on specialized computer chips they say that can interpret contextual data around a humanoid robot and tell it to stop working when conditions are deemed unsafe.
For now, at least, maybe it’s best to leave the fruit slicing to the humans.
In a world where tech leaders are racing to build autonomous machines that can do almost anything, one hobbyist is going the complete opposite route: an over-engineered robot that does exactly one thing and one thing only.
T-Kuhn’s Hexa-Bouncer only bounces a ping-pong ball, but looks like it’s capable of far more. The six-armed, metal machine was first designed in a simulated environment and then brought to life with a CNC milling machine. Custom-built joints let it rise, fall, and tilt, and even bob up and down while tracing a circle. A pair of cameras, set 120 degrees apart, track the ball’s position and send that data to a computer over USB. The computer then builds a 3D picture of the scene, then calculates when the paddle should strike the ball and at what angle. All of this happens in a fraction of a second.
Hexa-Bouncer keeps balls in the air remarkably well. In a video, Kuhn holds a ping-pong ball above his creation, and the machine moves up and down in sync with his hand. When he drops the ball, the robot immediately reacts, striking it back into the air and keeping it aloft bounce after bounce, all the while making small adjustments. After a few more bounces, the device eases off its swing and the ball starts bouncing lower and lower. It’s unclear exactly how long Hexa-Bouncer can keep the ball in the air, it never misses—at least in the video posted online..
Kuhn describes his ball juggler as the latest in a series of, “useless but mildly interesting ball bouncing machines.” Prior to this, he released his Octo-Bouncer, a similar design that keeps a ping-pong ball bouncing in the air using a four-sided flat plate. Some might say all this over-engineered ball bouncing is unnecessary, but Kuhn would disagree. For him, solving the design puzzle is a journey in itself.
“Building a thing like this is like bashing your way through a jungle,” he said. “It’s an adventure. You’re not completely sure if you can make it through or not, and that’s exactly why you do it.”
Agility Robotics has debuted its first humanoid robot engineered to work safely near humans without risking harm to flesh-and-blood coworkers. Such safety features could unlock many more opportunities to use such robots inside warehouses and automotive factories—all without requiring isolated robot work cells and physical separation barriers.
When Agility’s new Digit 5 robot detects a person at a distance, it can autonomously take precautions, like moving to avoid the person or standing still so the person can pass by without getting closer. If a person is getting into close proximity with Digit 5, the robot can even choose to squat and assume a seated position.
“The robot was designed with a complex safe motion system that can take a variety of different mitigations depending on exactly what sort of human presence is detected,” Pras Velagapudi, chief technology officer at Agility, told Ars.
Será mais fácil ensinar um robot a andar pela casa ou simplesmente construir uma casa à volta de um robot? A startup japonesa MW aposta na segunda hipótese.
China leads the world in churning out humanoid robots and four-legged robot dogs that also happen to be the most affordable on the market—and Unitree Robotics’ founder Wang Xingxing is arguably one of the people most responsible for that Chinese lead.
The introverted founder, who prominently appeared at a 2025 business symposium hosted by Chinese President Xi Jinping, became phenomenally wealthy after Unitree launched an initial public offering on the Shanghai Stock Exchange STAR Market on August 19. But extensive reporting by Beijing-based Caijing Magazine suggests Unitree’s success so far has been driven by Wang’s extreme micromanagement leadership style—an approach that may be more suited to a small startup than a fast-growing robotics company.
Caijing’s interviews with Unitree employees and investors paint a picture of Wang as someone who personally decides nearly every aspect of corporate strategy or product design, including the colors of materials and lengths of individual screws. The Caijing Magazine feature published on August 31, titled “The King of Unitree,” was translated into English by ChinaTalk, a US-based think tank and media organization, on September 10.
China leads the world in churning out humanoid robots and four-legged robot dogs that also happen to be the most affordable on the market—and Unitree Robotics’ founder Wang Xingxing is arguably one of the people most responsible for that Chinese lead.
The introverted founder, who prominently appeared at a 2025 business symposium hosted by Chinese President Xi Jinping, became phenomenally wealthy after Unitree launched an initial public offering on the Shanghai Stock Exchange STAR Market on August 19. But extensive reporting by Beijing-based Caijing Magazine suggests Unitree’s success so far has been driven by Wang’s extreme micromanagement leadership style—an approach that may be more suited to a small startup than a fast-growing robotics company.
Caijing’s interviews with Unitree employees and investors paint a picture of Wang as someone who personally decides nearly every aspect of corporate strategy or product design, including the colors of materials and lengths of individual screws. The Caijing Magazine feature published on August 31, titled “The King of Unitree,” was translated into English by ChinaTalk, a US-based think tank and media organization, on September 10.
On a sunny morning in June, I walked to work with a quadruped robot beside me. I’ve never gotten more attention from strangers.
A bunch of people snapped pictures of my robot dog. Several people asked me questions. Was it mine? (Yes.) Did I build it? (No.) Was it being used for surveillance? (No.)
Biological dogs kept a safe distance from my mechanical companion. Some growled or barked at it.
Stephen Joseph theatre, Scarborough The playwright explores our relationship with technology, asks if robots have souls and even questions the value of art in this sci-fi courtroom drama
For play number 92, Alan Ayckbourn returns to his preoccupation of exploring where we as a species are heading through a science-fiction lens. He goes back to the future. In The Trial of Romeo Oscar, the playwright famed for his domestic comedies looks to the year 2194 when humans have spread about the galaxy and androids live alongside us. One of them, RSDO83, is on trial for the murder of a teenage girl in his charge.
This is Ayckbourn philosophising about the future of humanity as we sit at an obvious inflection point, asking what this moment in our relationship with technology means and where it will lead us. The questions running through the play are manifold, complex and echo those asked by sci-fi writers for decades: can robots hurt humans? Do they have souls? He is also questioning the value of art and even its purpose.
Much like the earworm “The Song That Doesn’t End” from Shari Lewis’ Lamb Chops’ Play-Along, a new soft robot does not stop…curling up and jumping. A team of mechanical engineers from North Carolina State University say that they’ve created the bracelet-shaped bot that moves indefinitely when it’s exposed to infrared light.
Called a Ring Leaper, the teardrop-shaped machine is made of a liquid crystal elastomer ring with a more rigid aluminum V-shaped tail at the end. The ribbony-soft part contracts and tightens when exposed to the light. It keeps curling in, getting tighter and tighter (and building up more energy) until it reaches a “critical threshold” and folds onto the metal part. At that point, the V-shaped aluminum aggressively snaps downward, smacking the ground and launching the robot forward or into the air.
But Ring Leaper doesn’t quit after that. As soon as the bottom part snaps down, the ribbon part of the body will start tightening around itself all over again. This team of researchers at North Carolina State University says this results in a new kind of robot system that “autonomously stores, releases, and recycles elastic energy.”
Their mechanical jumping bean can leap either straight up or forward, depending on the angle of the V-shaped bottom part. And it has some hops. The tiny Ring Leaper can achieve a vertical leap exceeding 80 times its own body height. That would be the equivalent of LeBron James (six foot nine inches) leaping 540 feet into the air. That’s taller than a 50-story building. The team recently published their findings in the journal PNAS.
Ring Leaper can hop over hurdles and swim like a frog
Ring Leaper wasn’t limited to a sterile lab environment. It can tackle small obstacles like grass, rocks, and mulch. It can even swim… sort of. One of the tests involved placing Ring Leaper into a shallow tub of water before shining infrared light on it. On cue, it started its curling-and-jumping routine, but this time it would slightly pop out of the water and gradually wade forward like a frog. Ring Leaper was also able to clear a pair of tiny hurdles, even after slightly tumbling over the first one. When tasked with jumping headfirst into a tall wall, the robot dutifully complies, then whips backward and continues jumping back in the direction it came from, in a kind of mirroring effect. None of this, it’s worth saying, is really programmed or coded instruction. Ring Leaper just jumps.
Controlling Ring Leaper’s mode of movement is also pretty low-tech. The team can dictate whether it jumps straight up, propels itself forward, or goes somewhere in between by adjusting the angle of the V-shaped aluminum tube at its base. A V-shaped angle wider than 120 degrees will result in a crawling motion. Reduce that angle to 90 degrees, and it will leap forward like a frog. A narrower opening below 50 degrees results in a dramatic vertical jump. Each of those movements is also determined in part by the strength of the infrared light shining at it. More light equals more power.
“We could also improve the robot’s jumping distance by adding a small amount of weight to the rounded end of the teardrop,” Fangjie Qi, a North Carolina State postdoctoral researcher and study co-author said in a statement. “This changes the robot’s center of mass and leads to a more powerful and stable forward motion – much like a swimmer leaning forward as they dive off the starting block.”
So what’s this forever jumping robot actually useful for? Well, for now, nothing. Like a great deal of important scientific research conducted over many decades, the team essentially made it because they could, with no obvious commercial or practical application. That still doesn’t mean it’s completely useless. The proof-of-concept robot marks an advancement in self-resetting jumping, which could help other engineering teams who are looking to integrate these findings into future robotic navigation methods.
The deepest parts of our ocean are more like the fare reaches of space than planet Earth. We never really know what we’ll find down there. A team from the National Oceanographic and Atmospheric Administration’s (NOAA) Ocean Exploration team experienced that first hand when they spotted a very tiny fish in the Northeast Canyons and Seamounts Marine National Monument in the Atlantic Ocean off the coast of New England.
The team estimated that it measured only five millimeters (0.19 inches) to one centimeter (0.4 inches). It is still unclear what species this fish is, but scientists speculated that it could be a juvenile toadfish (family Batrachoidida) aka the sea toad. They have a sad-looking “Monday Face” and dwell on the sea floor to hunt whatever comes within their reach.
A glass sponge amongst polymetallic nodules. It was one of several varieties of sponges seen during the final dive of the 2026 Cook Islands ROV Exploration cruise. Image: NOAA Ocean Exploration, 2026 Cook Islands ROV Exploration.
Currently, NOAA is partnering with Nautilus Live to explore the deep sea near Wake Island. The island is about halfway between the Hawaiian and Mariana Islands, and is one of the most isolated land masses on Earth. The 30-day expedition will utilize remotely operated vehicles (ROV), mapping technology, and the E/V Nautilus to collect baseline data about the 157,236 square-mile area and the marine life that lives there.
The expedition could find potential maritime heritage sites associated with the Battle of Wake Island. From December 8 to 23, 1941, a small force of United States Marines and civilian defenders fought off parts of the Imperial Japanese Navy. The Japanese Navy ultimately took Wake Island, but at great cost. According to the National World War II Museum, the Japanese Navy ultimately sacrificed two destroyers, one submarine, 21 aircraft, and 900 to 1,000 lives just to conquer Wake Island. Despite the loss, fewer than 100 Americans died during the 16-day siege, including 46 Marines and three sailors.
The Wake Island expedition is scheduled to end on September 18.
Meta is testing robots that can plug in cables, reset servers, and handle other tasks inside its data centers, according to several current and former workers familiar with the projects. The ongoing effort, which has not been previously reported, may eventually allow Meta to operate its rapidly expanding data center footprint with fewer humans, keeping labor costs in check as its spending on AI infrastructure soars.
Meta is using robots and related hardware from several different vendors, including Watney Robotics, Kinova, and ABB, according to the same workers, who asked to remain anonymous because they weren’t authorized to speak to the media. Kinova and ABB declined to comment. Watney didn’t respond to requests for comment.
In one experiment, Meta is evaluating whether a Kinova Gen3 robotic arm could be used for power cycling or cutting off electricity to servers. The company is also testing a different robot to swap networking cables. One Meta data center worker estimates that if it’s successful, the bot could replace up to 80 percent of some people’s workloads. “We thought those of us performing the physical tasks were safe for a while, but not anymore,” says the worker. “It’s coming for us all, unfortunately.”
A leading US distributor of China’s most popular humanoid robots and robot dogs now plans to produce its own robots at a facility on Long Island, New York. The business pivot comes amid the US government’s growing crackdown on foreign-made robots.
The company, RoboStore, had previously become the main North American distributor for humanoid robots and quadruped robots from Unitree Robotics, a Chinese robotics company that produces some of the most affordable and popular humanoid robot models. Universities like Harvard and MIT, along with tech companies like Amazon and Nvidia, purchased such robots made in China through RoboStore for their own research and development purposes.
“We’ve sold over 1,500 robots and deployed them and worked with customers like Cisco, OpenAI, Nvidia, and over 150 universities,” Teddy Haggerty, founder and CEO of RoboStore, told Ars.
O robot Threehalves tem uma estranha forma de centauro e está equipado com uma motosserra, construído para penetrar em locais onde os humanos não conseguem, para resgatar vítimas de incidentes.
Mitsubishi is joining the rapidly growing ranks of robot builders, with plans to build up to 1,000 of its Highlanders humanoid robots per month at one of the automaker’s ICE engine plants in Kyoto – and if the plan comes together, the automaker could begin mass-producing its new worker-bots as early as next year.
Researchers at the University of Queensland have developed a new type of sensor that can measure how hard muscles are working without needles, surgery or other invasive procedures. The breakthrough could help create smarter robotic limbs, wearable exoskeletons and rehabilitation devices that respond more naturally to a person’s movements. The study was published in Science […]
Login
