"Blue Book on Internet Advertising Governance in the AIGC Era" was released, advocating "governing AI with AI"

China News Service, Beijing, October 24 (Yin Qianyun, Tao Siyue) On October 24, during the 33rd China International Advertising Festival, the "Blue Book on Internet Advertising Governance in the AIGC Era" was officially released.

At present, the other side of the efficiency improvement brought by AIGC technology is the escalation of risks: black and gray products use AI to generate high-fidelity materials in batches and forge authoritative endorsements, which are difficult to deal with only by manual review. In the face of these new risks, the blue book provides a governance method throughout the entire process of advertising: it divides risks into four categories: content violations, technical confrontation, transaction links, and new scenarios, and makes suggestions from eight aspects.

For example, the Blue Book presents a practical approach to “governing AI with AI”: the first step in governance is to see the risks clearly. Using a large multi-modal model to view pictures, read words, and listen to sounds at the same time, you can not only understand the surface meaning, but also determine whether there is a violation based on the context. Xiao Tao, head of the commercial security content security strategy of Jumbo Engine, said that relying on the intelligent review system, 90% of the materials can be reviewed within 10 minutes at the fastest, increasing efficiency by 75%.

In addition, the AI ​​application platform also incorporates user feedback into management and provides special protection for the elderly and minors. The elderly can provide evidence and go directly to manual customer service with "zero threshold". Parents can use standardized templates to report problems, and some problems can be solved on the same day.

After the unification of standards, the platform maintains regular risk monitoring and inspections, and studies and determines risk trends in advance around high-risk issues; for new variants that have changed their appearance, it quickly identifies and timely updates interception strategies to promote identification and disposal; at the same time, it continues to iterate the identification model and review rules, and severely punishes malicious counterfeiters and repeat offenders until they are shut down and removed. Those suspected of violating the law are reported to the relevant departments in accordance with the law or cooperate with the investigation and punishment.

Liu Xiaochun, director of the Internet Legal Research Center of the University of Chinese Academy of Social Sciences, said that governance is not to constrain innovation, but to build a solid foundation of trust for AIGC advertising. She believes that AI cannot create a vacuum of responsibility. Advertisers bear the first responsibility for authenticity and cannot be exempted from liability on the grounds that "content is generated by AI". Platforms bear the responsibility of checking that matches their technical capabilities, and all parties have their place.