IT之家 reported on October 26 that, according to CCTV Finance, at the recently concluded 2026 Financial Street Forum Annual Meeting,the nation's first comprehensive artificial intelligence application risk protection plan was officially released.Insurance products provide coverage respectively for losses caused to third parties by artificial intelligence services, and for the artificial intelligence provider's own expense losses,filling a gap in the domestic AI insurance market。
IT之家 noted that AI is taking over more and more tasks: making phone calls, serving as customer service, analyzing agricultural product prices, writing copy, and editing videos. But once it “says something wrong” and commits infringement, or “miscalculates” and causes losses, who is responsible?
In September this year, the Supreme People's Court issued the “Opinions on Adjudicating Artificial Intelligence-Related Dispute Cases in Accordance with the Law,” clarifying key rules:Generative AI infringement generally applies fault-based liability, and “the content was generated by AI” cannot be used as a defense for exemption from liability.In other words, whoever is at fault bears the responsibility; you cannot shift the blame with a line like “the machine did it.” Even if the service provider only supplied model capabilities, if it failed to fulfill its duty of care, it may also have to bear compensation liability.
According to reports, this type of insurance coverscontent generated by the insured during the provision of generative artificial intelligence services due to negligence or oversight,such as text, images, audio, video, etc., that infringes upon the legitimate rights and interests of third parties, or misleads third parties, and directly causes personal injury or property damage.
Of course, insurers do not underwrite blindly; they will first look at what scenarios the AI is used in, what content it generates, and how much data it processes daily before setting premiums. If something really goes wrong, it is not an “automatic full payout”—it depends on whether the enterprise should bear legal liability and how the contract stipulates it.。
In addition to coverage for losses caused by AI to third parties, there is another type of insuranceaimed at the AI service provider's own losses.When model performance deviates, expenses such as repairs, refunds, and appraisal are borne by the insurance company as stipulated in the contract.
Before underwriting, insurance companies conduct a comprehensive assessment of risk factors such as model parameters, application scenarios, and actual performance, implementing differentiated pricing. These productsare mainly aimed at AI large model R&D enterprises, AI application developers, smart hardware manufacturers, and AI solution providers.Currently, similar insurance products have begun to extend to scenarios such as office agents, language large models, and humanoid robots.