China’s National Information Security Standardization Technical Committee has released a draft document outlining rules for determining problematic generative AI models. The document provides criteria for banning data sources, demands diversification of training materials, and sets requirements for hiring moderators. It also outlines what constitutes prohibited content and addresses the need for more subtle censorship. While these standards are not laws, they may influence future AI regulations, and Chinese tech companies have a significant role in shaping them. The detailed requirements may have a significant impact on the global AI industry.
China’s New Plan for Assessing the Safety of Generative AI
In July, China passed a law on generative AI, raising questions about how the country’s censorship machine would adapt in the AI era. Last week, the National Information Security Standardization Technical Committee released a draft document outlining detailed rules for determining whether a generative AI model is problematic. These rules provide companies with clear criteria to comply with the generative AI regulations, which were previously vague. The document also addresses concerns regarding safety risks and content censorship.
Practical Solutions:
- Training: Companies must diversify the corpora used to train AI models and assess the quality of training materials. They should randomly sample 4,000 pieces of data from a source and blacklist it if over 5% of the data is considered “illegal and negative information.”
- Moderation: AI companies should hire moderators to improve the quality of generated content promptly. The size of the moderator team should match the size of the service.
- Prohibited Content: Companies must select keywords for flagging unsafe or banned content. Political and discriminative content categories have specific keyword requirements. Companies need to provide test prompts and ensure that fewer than 10% of generated responses break the rules.
- Sophisticated Censorship: AI models should not make their moderation or censorship too obvious. Companies are asked to find acceptable prompts related to sensitive topics and ensure that AI models refuse to answer fewer than 5% of them.
It’s important to note that these standards are not laws and compliance does not carry penalties. However, they are expected to influence future regulations, and companies are likely to treat them as binding. The TC260 standards are shaped with input from experts hired by tech companies, including influential companies like Huawei, Alibaba, and Tencent. The impact of these standards on the global AI industry cannot be underestimated.
To stay competitive and leverage AI in your company, consider the following steps:
- Identify Automation Opportunities: Locate key customer interaction points that can benefit from AI.
- Define KPIs: Ensure your AI endeavors have measurable impacts on business outcomes.
- Select an AI Solution: Choose tools that align with your needs and provide customization.
- Implement Gradually: Start with a pilot, gather data, and expand AI usage judiciously.
For advice on AI KPI management and continuous insights into leveraging AI, connect with us at hello@itinai.com. To explore practical AI solutions for automating customer engagement and managing interactions, visit itinai.com/aisalesbot.
