Recent discussions about ordinary people earning 3,500 yuan in three days through "selling their faces" and AI image licensing have sparked widespread debate. From celebrities testing the waters early on to platforms recently opening up to the general public, what does this AI image licensing business really entail? What are its positive implications and potential risks? If ordinary individuals want to try it, how should they protect themselves? Let's explore these questions together.
How does AI image licensing work, and does the process differ between celebrities and ordinary people?
Li Ziqiang, AI business director at Ciwen Media Co., Ltd., explains that the primary method involves registering on a relevant platform, uploading a personal image, and signing an agreement. Currently, AI digital image licensing is categorized into two types: ordinary people and professional actors. Ordinary people are paid based on the duration of their appearance in the finished video, with rates ranging from several hundred to a thousand yuan per minute. Professional actors negotiate individualized contracts, with prices tiered by usage scope and duration, ranging from thousands to tens of thousands of yuan.
For ordinary people, the most direct benefit is realizing the dream of appearing on screen at an extremely low cost, turning one's personal image into a sustainable digital asset. For the industry, it significantly enriches content supply and pushes production towards more accessible and intelligent upgrades. Xu Xiaodan, deputy director of the Culture, Tourism, Media and Sports Committee of the Beijing Lawyers Association and a partner at Beijing Yuankun Law Firm, points out that the right to one's likeness is a personality right, which cannot be transferred or bought out but can only be licensed for use. Facial features and voiceprints are sensitive biometric information that requires separate explicit consent. Celebrities and ordinary people are completely equal in legal rights, with differences lying in commercial value, social influence, and legal recourse capabilities. Ordinary people often lack professional teams to review contracts, making them more vulnerable to risks during agreement negotiations.
What are the positive aspects and potential risks associated with AI image licensing?
Li Ziqiang believes that AI and human co-creation can continuously add vitality to the industry, but they are complementary rather than a substitute for each other. Ordinary people can use AI avatars for preserving personal videos and family memories, as well as participating in AI films, creating short videos, and virtual live streaming for small amounts of monetization. Celebrities can break through age and schedule limitations, extend their careers, revitalize classic IPs, and also benefit from the cost reduction and efficiency gains in film reshoots and advertising production. However, opportunities coexist with risks.
Xu Xiaodan analyzes that some unscrupulous platforms set low-priced, unequal licensing agreements, with clauses hidden for "perpetual authorization," "allowing sub-licensing," or "use for model training." Once signed, facial data may be repeatedly transferred. Furthermore, ordinary people face difficulties in providing evidence for rights protection. After multiple iterations of AI models, proving that a character originates from one's own facial modeling involves high costs for technical appraisal and challenges related to dispersed responsible parties.
How should ordinary individuals mitigate risks if they wish to try it? What improvements are needed in the industry?
Xu Xiaodan advises that the public must choose compliant platforms with verifiable credibility and thoroughly understand the contract details before signing. It is crucial to reject vague clauses such as "perpetual authorization," "unlimited sub-licensing," or "using facial data for large model training." The licensing period and specific projects should be strictly defined. Separate consent terms must be signed for facial features and voiceprints. After the contract expires, require the complete deletion of original materials, training models, and derivative data, and retain proof of destruction. On the platform and corporate side, Xu Xiaodan suggests encrypted storage of biometric data, auditable data destruction upon license termination, mandatory clear labeling of all AI-generated content as "AI Generated," and calling on industry associations to quickly establish standards for AI digital human licensing. This would improve processes for authorization, data destruction, and dispute resolution, enhancing legal risk awareness among market players.
Li Ziqiang states that the industry is actively working to establish a comprehensive system for likeness rights compliance and authorization confirmation, promoting tiered licensing and transparent revenue-sharing mechanisms to lay a solid foundation for the long-term healthy development of the sector.