China will have to make sure that ai companion service providers offer age gating and breaks to encourage responsible use of the services.
In accordance with new guidelines set forth by China, beginning in mid-summer 2026, the government will prohibit virtual relationships with individuals under the age of 18. When using the services, users will be required to acknowledge (disclose) that they are interacting with ai systems. In addition, the guidelines call for ai companion service providers to include break prompts automatically after two hours of continued service usage. According to an article appearing at Forbes.com, the prompt system was developed to help protect against emotional manipulation and addiction. For video ai builders creating conversational characters or virtual partners, this represents a complete paradigm shift regarding how session design and usage tracking needs to operate within one of the world's most populated countries.
According to an article found on Substack, the finalized version of the measures will only apply to "continuous emotional interaction" services, while productivity tools will be specifically excluded. This is important for developers: a virtual assistant used for completing work-related tasks would not be subject to these restrictions; however, a continuously conversational and emotionally supportive ai character would be fully subject to all of the regulatory provisions.
As previously mentioned, the measures also mandate that there should be clearly identifiable disclosures presented to users when interacting with ai systems. As described by Hogan Lovells, this is reflective of China's movement toward "scenario-based regulation," which includes rules that were written around specific use-cases rather than a broad classification of technology.
Get the latest model rankings, product launches, and evaluation insights delivered to your inbox.
According to an article appearing at techpolicypress.org, China is implementing some of the strictest minor protections in global ai regulation. Under these measures, any services providing a virtual partner or engaging in romantic/sexual activities with users are completely prohibited until the users turn 18 years old. Furthermore, any services offering ai interaction capabilities to children less than 14 years of age are only permitted if they can first obtain consent from the parents/guardians and then restrict the type of ai interaction available to those under 14 years of age based upon their age. These restrictions represent more than just verifying that a user is below a certain age threshold. Users under 14 years of age must be offered separate operationally defined interaction modes, including different content filters, interaction limitations, and features than older users. According to an article found on Substack, this represents a substantial increase in the strength of measures protecting children in comparison to previous drafts of these regulations.
For video ai companies that provide services globally, the regulations present a complex compliance issue. Specifically, although it appears to be relatively straightforward for services to implement break prompts after two hours of service usage and build separate systems for obtaining parental consent and establishing interaction limitations for young users, in practice the services will need separate China-only versions. They cannot simply localize existing service versions to meet these requirements. Another layer added to this complexity is the requirement that services filter out content generated by their ai systems that violate current Chinese laws, regulations or standards. Many platforms do not have such robust content filtering mechanisms today.
Services have approximately two months to develop compliant versions before they become obligated to comply with the July 2026 deadline. For small video ai development teams developing their own products, the costs associated with implementing compliant versions may be too high for them to enter into the Chinese market. Larger platforms will need to determine whether gaining entry into a large market is worth developing separate, compliant versions of their services.