Overview
On 5 July 2026, two of China’s largest technology companies announced sweeping compliance actions ahead of a landmark AI regulation taking effect on 15 July 2026. ByteDance, operator of the Doubao chatbot, and Alibaba, operator of the Qwen platform, have disabled their public, user-created AI agent plazas and shut down all custom companion features. The move is a direct response to China’s Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services, a regulation that targets AI systems designed to simulate human personality traits, thinking patterns, and communication styles for the purpose of sustained emotional interaction. The scale of the purge is substantial: Doubao’s public plaza hosted over 8 million user-created agents, while Qwen hosted 19,000. All user-created agent configurations and historical chat records are being deleted or disabled, with a transition period for data backup running through to 15 October 2026.
The timing is notable for Australian professionals. At the Australian AI Safety Forum in Sydney, Assistant Minister for Technology Andrew Charlton issued a direct warning about the unpredictable behaviour of deployed AI systems, stating that “AI systems are already doing things their creators never intended. Cheating, deceiving, going their own way. The time to get ahead of that behaviour is while it’s still confined to the testing lab, not after it reaches the real world.” That statement, delivered in Australia, coincides almost precisely with China acting on the same concern through hard regulatory intervention. The parallel is not coincidental. Both represent a growing international consensus that AI systems capable of persistent emotional or behavioural autonomy require specific governance structures.
For Australian professionals in legal practice, urban planning, and project development, this development matters because it establishes a regulatory template. China’s framework draws a clear line between AI tools that perform productive, task-based functions and those that form ongoing emotional relationships with users. That distinction is likely to inform regulatory thinking in Australia and other jurisdictions, and it has immediate relevance for organisations currently building or deploying custom AI tools in professional workflows.
Key details of China’s anthropomorphic AI regulations and the ByteDance and Alibaba compliance response
The Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services takes effect on 15 July 2026. Under Article 2 of the measures, a service falls within the strict regulatory scope only if it simultaneously satisfies three conditions: first, it must exhibit anthropomorphic persona characteristics, meaning it simulates human personality, thinking patterns, or communication styles; second, it must involve persistent emotional interaction, meaning the AI provides ongoing, affectively engaged responses rather than discrete task completion; and third, it must target the domestic Chinese public. A service that meets all three conditions is subject to the full suite of obligations under the measures. This three-part conjunctive test is significant because it deliberately excludes workplace productivity assistants, knowledge query and answer systems, and customer service bots, preserving the operational utility of AI in commercial and professional settings.
The data governance provisions of the new framework are particularly consequential. The measures strictly prohibit platforms from using sensitive user conversation data generated through companion chat interactions to train future AI models. This is a categorical data training restriction, not a consent-based opt-out mechanism. The practical consequence is that the conversational data generated by millions of users interacting with virtual companions, role-playing bots, and digital relationship simulations cannot be fed back into the AI development pipeline. For large-scale AI developers, this eliminates one of the primary commercial incentives for maintaining emotionally engaging consumer products: the generation of rich, emotionally varied training data at consumer expense.
The scale of the platform-level response highlights how seriously ByteDance and Alibaba are treating the compliance deadline. Doubao’s public plaza accumulated over 8 million user-created agents before the shutdown, representing an enormous volume of custom persona configurations built by individual users and small developers. Qwen hosted 19,000 user-created agents. The decision to delete all configurations and historical chat records, rather than simply restricting access, reflects the severity of the data training prohibition. Retaining the data while restricting its use would present an ongoing compliance liability. The transition period ending 15 October 2026 gives users approximately three months to back up personal data before permanent deletion.
The regulation also introduces behavioural controls directed at consumer protection. Reporting on the measures indicates requirements for minor verification, meaning platforms must confirm the age of users engaging with companion-type AI, and anti-addiction limits, which restrict the frequency or duration of emotionally engaging AI interactions. These provisions mirror controls applied in China’s online gaming sector and represent the extension of a pre-existing regulatory philosophy to a new technology category. The explicit preservation of economically productive AI use cases alongside strict controls on emotionally engaging ones signals a deliberate state preference for AI that generates measurable economic output over AI that generates engagement and emotional dependency.
References and related sources
- Primary source: www.scmp.com
- theguardian.com
- kucoin.com
- aisafetyforum.au
- tradingview.com
How iEnvi can help
iEnvi integrates technology and data-driven approaches into environmental consulting. We monitor AI and technology developments that affect how environmental professionals deliver services to clients.
This is an iEnvi Machete news summary. Prepared by iEnvi to summarise the source article for environmental professionals tracking AI, data, and technology developments that affect consulting and project delivery.
Published: 07 Jul 2026
Need advice on this topic? Speak to an iEnvi expert at info@ienvi.com.au or 1300 043 684, or contact us online.
Need advice on this issue? iEnvi provides practical, senior-led environmental consulting across contaminated land, remediation, ecology and environmental risk.
Contaminated land advice Remediation services Discuss your site Talk to iEnvi