en.Wedoany.com Reported - Two major AI products, Doubao and Qianwen, recently announced that they will officially discontinue their platform AI agent functions starting July 15, 2026, requiring users to back up the configurations and historical conversation data of their custom-built agents in advance. The effective date of this adjustment coincides with the official implementation date of the "Artificial Intelligence Anthropomorphic Interactive Service Management Interim Measures" jointly issued by five ministries including the Cyberspace Administration of China, marking the entry of the general large model industry into a compliance rectification phase.
Doubao platform issued a notice stating that the AI agent function will be discontinued on July 15 due to product adjustments. Subsequently, users can view relevant materials and conversation records within a certain period, and it is recommended to back up via screenshots or export. Starting October 15, data not retained will be cleaned up in accordance with the privacy policy. ByteDance's Catbox APP supports the creation and use of new AI agents, which can accommodate some user needs.
On the same day, Qianwen (Tongyi Qianwen) also pushed a notification to users about the discontinuation of its AI agent service, stating that the function will cease service on July 15, after which users will no longer be able to access related configurations and conversation records. Users can update to Qianwen APP version 6.12.9 or above, export conversation records on the AI agent page, and view custom agent parameters. It is recommended to save key content in advance through copying, screenshots, or text export.
This adjustment only applies to AI agents independently created by users; the platform's official built-in standardized AI Q&A, office, and creative tools are not affected. At the business level, ByteDance will migrate anthropomorphic intelligent interaction services to dedicated vertical products, while Tongyi Qianwen will spin off compliant AI agent services from the main site for independent planning and launch.
According to public data, the Doubao platform had 8 million AI agents in 2024, and after nearly two years of development, the actual scale is even larger, making personalized fine-tuning review for each agent highly challenging. Experts point out that although large models themselves have completed filing, AI agents built on top of them may still pose security risks. Developers could use external knowledge bases and prompts to bypass underlying security restrictions, leading to content and data security hazards. On June 26 this year, the Shanghai Cyberspace Administration announced the first-phase results of the "Qinglang·Rectification of AI Application Chaos" special action, with a total of over 14,000 non-compliant AI agents removed, involving platforms such as Xiyu, Xiaohongshu, and Bilibili, primarily targeting violations like "one-click undressing," gambling, and AI-managed operations.
The timing of the simultaneous adjustment of AI agents by the two major platforms coincides with the official implementation date of the "Artificial Intelligence Anthropomorphic Interactive Service Management Interim Measures" by five ministries. The measures, jointly issued by five departments including the Cyberspace Administration of China, take effect on July 15, 2026, requiring anthropomorphic AI services to complete role filing, implement full-chain real-time review, enforce tiered management for minors, and achieve risk traceability. General large model main sites host a large number of user-generated AI agents, with complex scenarios and vast quantities, making it difficult to meet full-coverage review standards. Industry insiders believe that this simultaneous rectification by leading platforms marks the starting point of industry standardization, and other mainstream large model platforms may follow with similar adjustments.
