Wuhan court makes AI production costs part of copyright damages

The ruling, first reported by the National Law Review and picked up by The Decoder on 28 September, concerns a one-hour short drama produced with AI tools in early 2026 and published on platforms including WeChat. A day later, another company copied the work, retitled it and ran ads against it.

The Wuhan court treated the drama as a protectable audiovisual work. It reasoned that employees made creative decisions at every stage: the script, the prompt design, the selection of AI outputs and the final edit. The AI served as a tool. Alongside the AI-specific costs, the court weighed runtime, distribution reach and the duration of the infringement.

The award is small, but the precedent is significant. The decision extends a line of Chinese copyright rulings on machine-generated output that began in 2023, when a Beijing court granted protection to an AI-generated image. The court also advised creators to keep scripts, prompt drafts and project files. The paper trail is now evidence.

Factories, not courtrooms, set the pace

The same week, the industrial base for the robots those models are meant to drive got a hard number. The International Federation of Robotics said on 27 September that the global operational stock of industrial robots rose 9% to a record 5 million units in 2025. Annual installations jumped 11%, with more than 600,000 new units shipped. "The new mark of five million robots operational in factories worldwide is more than double the number seven years ago," IFR president Jane Heffner said in the release. The Robot Report carried the same figures on 24 September.

China accounted for 59% of global deployments, installing 354,000 units in 2025, up 20% year on year. The United States passed Japan to become the second-largest market, with installations up 12% to almost 38,500 units. Japan fell 19% to 36,219 units. The IFR forecasts installations rising 9% to 655,000 units in 2026 and reaching 806,000 in 2029.

Those are the boring robots: arms bolted to a line, not androids. The humanoid push is a different, messier bet, and the reporting this month has been mostly about the mess.

Tesla's hands, Toyota's jigs

Ars Technica reported on 26 September that Tesla has scaled Optimus production to hundreds of robots per week. The target is above 1,000 per week by the end of 2026, citing extensive reporting by The Information. The blockers are specific. Optimus hands and forearms contain more than 100 small components that workers assemble by hand. Touch sensors have proven unreliable enough that Tesla developed a glove-like sensor layer that can be swapped without replacing the whole hand. Some workers objected to wearing motion-capture suits to train machines they "knew the robots were designed to eventually replace." Tesla has moved the data collection to dedicated teams and training hubs, according to The Information.

Toyota is further along in framing, if not in deployment. Ars Technica reported on 23 September that the group plans to invest $6.42 billion a year starting in 2028. It aims at 150,000 robots in its own plants and another 250,000 at group facilities. Workers are already teaching ELEY humanoids tasks requiring precise hand movements by wearing "jigs" modelled on the robots' fingers, Nikkei Asia reported. Executive vice president Hiroki Nakajima said the goal is "a world where robots coexist with humans, rather than replacing them."

Agility Robotics, meanwhile, is hedging its own form factor. CEO Jonathan Hurst told The Robot Report on 25 September that the company is "exploring a range of robot designs and form factors, including wheels," after a wheeled robot appeared in the Digit 5 launch video. On timelines, he was blunt: "Too early to say." Agility is going public through a SPAC with Churchill Capital Corp. XI at a $2.5 billion valuation. It reported $1.8 million in 2025 revenue against a $140 million operating loss.

The AI layer is moving into the same rooms. On 27 September, The Decoder described HomeBody, a Stanford and Caltech system that lets a Unitree G1 clean an unfamiliar kitchen by calling GPT-6 Astra directly into a skill library, with no trained control layer in between. Latency, overheating finger servos and compute costs are listed as limitations.