In Wuhan, a taxi driver named Yao Xinnong watched his pay fall by about 40 percent after a fleet of driverless taxis called Apollo Go started picking up his regular customers. When the self driving cars broke down for months earlier this year after a mass malfunction, his income jumped right back up. That single detail tells you almost everything about how fast this is moving in China.
Apollo Go is run by the search engine company Baidu. It is not a small trial anymore. The service now runs in 22 cities worldwide, has completed more than 17 million paid rides, and grew its fully driverless orders by over 200 percent in one quarter compared with the year before. Baidu has said the Wuhan operation is close to breaking even. This is a business that went from curiosity to real competitor to human taxi drivers in about three years.
China's taxi drivers are part of a much bigger group. More than 200 million people in China work in gig style jobs: delivery, ride hailing, warehouse work, and similar roles. That is roughly a third of the country's workforce, and it grew that large partly because people who lost jobs in real estate and other struggling industries flowed into gig work as a fallback. Now the fallback itself is one of the first places AI is taking hold, which leaves fewer places for displaced workers to land.
The white collar side is moving too. A cinematographer in Beijing told us AI tools that made rough looking avatars a year ago can now produce convincing digital actors good enough to carry a full film, and his freelance rate has dropped to 40 percent of what it was in 2019.
Beijing's response is the part worth watching closest. A court in Hangzhou recently ruled that a company could not fire a quality assurance manager just because AI could do cheaper work, ordering the company to pay him around 38,000 dollars after it tried to demote him and cut his pay by 40 percent. The government has also rolled out new rules covering pay floors, working hour limits, and transparency around the algorithms that assign gig work, with a compliance deadline of 2027.
This is a shift from a year ago, when Beijing was pushing AI development after the success of the homegrown model DeepSeek with few limits. Officials are now talking about AI needing to support "shared prosperity" rather than just growth.
For any business outside China, this is a preview, not a foreign curiosity. China moves first because the state can push adoption fast and then step in with rules once the social strain shows up. If you are counting on AI to cut headcount costs, the Hangzhou ruling is a warning that regulators and courts elsewhere are likely to draw similar lines: AI can change how work gets done, but "AI is cheaper" is starting to look like a weak legal reason to end someone's job. The technology is not slowing down. The rules around using it to replace people are catching up faster than most companies expect.