Recently, Chinese AI companies Zhipu AI's "GLM-5.2" and Moonshot AI's "Kimi K3" have been drawing attention for their exceptional performance, but it has been pointed out that both services fall short of major global AI services in terms of privacy protection features.
A comparison of the privacy policies of Zhipu AI and Moonshot AI with other major AI services such as ChatGPT and Claude revealed the biggest difference in the "opt-out" function, which allows users to refuse the use of their data for AI training.
ChatGPT and Claude allow users to directly toggle off the use of conversation contents for model training within their service settings.
On the other hand, Zhipu AI and Kimi do not provide separate settings within their services, instead instructing users to submit individual requests via email asking for their data not to be used for model training.
This makes the process for users to refuse the utilization of their data for training relatively cumbersome.
In addition, Kimi's privacy policy explicitly states that it can collect "clipboard data," which is generated when users copy text or images.
Because sensitive information such as passwords, account numbers, and verification codes can be temporarily stored on the clipboard, experts point out that close attention needs to be paid to such data collection by Chinese AI companies.
Since both Zhipu AI and Moonshot AI are Chinese companies, there is also the possibility that user data could be provided to the Chinese government upon request under Chinese law.
The industry views that, much like the case last year when governments and companies around the world restricted the use of DeepSeek citing security concerns, privacy and security systems will become crucial competitive edges alongside performance in the generative AI competition moving forward.
Reported by Kim Minjeong | Video by Lee Yu-jin | Graphics by Yook Do-hyun | Produced by SBS Digital News
※
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.