Skip to content
Brief The AI-First Web ·

China's data regulator plans standards for AI token billing and quality

The National Data Administration wants common rules for how AI tokens are counted, billed and judged.

In brief
  • China's National Data Administration plans stronger standards for measuring, billing and judging the quality of AI tokens. Its director, Liu Liehong, said so in remarks posted on November 21.
  • The agency has not yet said what the standards will contain or when they will apply.

Tech in Asia reported that China's National Data Administration plans to strengthen standards for how AI tokens are measured, billed and judged for quality. The agency is the country's central data regulator. The report cites remarks by its director, Liu Liehong, posted on the agency's website on November 21. A token is a small unit of text that AI models count and charge by.

The agency said "token factories" are appearing. These firms bundle rented computing power, model supply and data services into token-production offerings. It also said it would widen the use of token services and work to stop low-quality copying as the industry grows. Tech in Asia noted that Alibaba, Baidu, ByteDance and iFlytek cut prices for large-model access in 2024, often charging per token or per call. The report gives no content for the standards, no start date, and does not say whether they will be mandatory.

Many teams buy AI model access by the token. The planned standards cover how tokens are counted, billed and judged, which are the basis for what buyers pay and how they compare vendors. Until the agency publishes details, the practical effect on providers and customers is unknown.

A WebPulse Brief: a short report of an important event, written by the WebPulse Newsroom with AI assistance and checked against the reporting below. When there is more to explain, we follow up with a full story. How we use AI.

Reporting: Tech in Asia.