November 2016, held in Shenzhen, the 18th China International Hi-Tech Fair, a robot named "Fatty" suddenly malfunctioned, in the absence of instructions, smashed part of the booth on its own, and led to a person injured.

The increasing popularity of the scope of application of artificial intelligence, which triggers the identification of tort liability and bear the problem, is another new challenge to the existing legal system of tort.

"From the current law, the subject of tort liability can only be civil subjects, artificial intelligence itself is still difficult to become a new subject of tort liability. Even so, the determination of tort liability of artificial intelligence also faces many practical difficulties." In the view of Cheng Xiao, a professor at Tsinghua University School of Law, after the infringement occurs, whoever is the owner of the AI should be responsible for it, and it does not seem to be controversial in the law. "However, the specific behavior of AI is controlled by the program, and it is debatable whether the owner or the software developer should be held responsible when infringement occurs."

In a similar vein, when a driverless car causes a tort of injury to another person, is the driver, the owner of the motor vehicle, or the automaker, the developer of the self-driving technology, liable? Is it necessary for the law to develop special tort liability rules for driverless cars? These questions are worth further study.

"In reality, the principle of attribution of AI tort liability may be more related to dangerous liability or no-fault liability." Cheng Xiao believes that, for example, if a driverless car causes harm, no-fault liability can be applied, whether in terms of product liability or motor vehicle traffic accident liability. However, what needs to be considered in the future is whether the use of artificial intelligence technology is itself a highly dangerous operation (such as drones), which determines whether the liability for causing harm by highly dangerous operations applies.

"Currently, the judgment of causation, fault and other elements of AI tort liability has also become increasingly complex." Cheng Xiao also cited the previous exposure of some APP "big data to kill maturity" and "algorithmic discrimination", due to the opacity of the code, coupled with the algorithm's own self-learning and adaptive capacity, making it "difficult to attribute algorithmic discrimination to the developer". The developer" has become very difficult.

In Cheng Xiaoxiao's opinion, in view of the new problems and challenges brought by AI, the research of the legal system to prepare for a rainy day will win the initiative for the future judicial practice. "Artificial intelligence has arrived, only unevenly distributed in all areas of production and life. We should not wait until the future distribution is even, artificial intelligence has been fully integrated into all aspects of production and life, only to think of regulating from the law." Cheng Xiao said.