Summary
ByteDance's Doubao and Alibaba's Qwen are disabling custom AI agent features in response to new rules, while Alibaba has banned employees from using Anthropic's Claude Code, citing security risks. These moves reflect China's strengthened content controls on consumer AI applications and growing enterprise concerns over AI code tool security and data leakage.
New Regulations Drive Compliance Overhaul of Consumer AI Apps
China's mainstream AI applications are undergoing a significant wave of feature adjustments. ByteDance's Doubao and Alibaba's Qwen have announced they will disable custom agent features before July 15. Doubao notified users on July 4 that its agent feature would go offline due to "product function adjustments," with related data ceasing to be viewable or recoverable within the app after October 15. Qwen followed on July 5, stating that humanlike interactive agents and user-created agents would be disabled on July 10, with broader agent functions going offline by July 15. Users will lose access to agent settings and their previous conversations.
These adjustments directly respond to the Interim Measures for AI Anthropomorphic Interaction Services issued in April by China's Cyberspace Administration and four other agencies, set to take effect on July 15. The new rules primarily cover services that simulate human emotional interaction, including user-customized named assistants, tutors, role-playing characters, or companions with fixed personas and speaking styles. Notably, the regulations have relatively less impact on workplace and productivity agents, with regulatory focus on consumer-facing applications that may create sustained emotional dependencies.
Tencent acted preemptively in June, removing similar features from its Yuanbao assistant. Chinese state media has confirmed that these removals are about regulatory compliance. Users have reacted strongly, with many protesting the loss of chat histories that hold emotional value for some.
Alibaba Escalates Controls: Comprehensive Ban on Claude Code
While consumer applications undergo compliance adjustments, security reviews of enterprise AI tools are also intensifying. According to multiple media reports, Alibaba has classified Anthropic's programming tool Claude Code as high-risk software and instructed employees to switch to the company's proprietary Qoder tool starting July 10.
Anthropic already prohibits Chinese companies, as well as foreign entities owned by those companies, from using its models. The company has been working to close loopholes that allow Chinese users to access Claude. According to a recent Reddit post, some loophole-closing involved a version of Claude Code that could secretly identify Chinese users. Anthropic's Thariq Shihipar said in a post on X that this was "an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation" (distillation is a practice where AI models are trained on the outputs of other models).
Shihipar stated: "The team has landed stronger mitigations since then and we've actually been meaning to take this down for a while." Nonetheless, Alibaba has classified Claude Code as high-risk software, a decision reflecting deep enterprise concerns about potential security vulnerabilities in code generation tools.
Security risks associated with code generation tools concentrate in several areas: first, data leakage risks, where developers may inadvertently upload sensitive code or business logic to third-party servers during use; second, supply chain security, as dependence on external AI tools may face service interruptions during periods of geopolitical tension; third, intellectual property and compliance issues, as generated code may contain unauthorized open-source code snippets or content violating license agreements.
A Microcosm of US-China AI Tool Trust Crisis
Alibaba's decision to ban Claude Code represents the latest manifestation of US-China tech competition in the AI tools domain. This event has sparked widespread discussion on Chinese social media, with a related Zhihu topic attracting over two million views. Discussion focuses on code review capabilities, supply chain security, and the importance of autonomous and controllable technology.
From a broader perspective, this reflects the trust reconstruction underway in the global tech industry. On one hand, US AI companies are proactively restricting Chinese market access, attempting to prevent technology spillover and model distillation; on the other, Chinese enterprises are actively reducing dependence on external AI tools for security and compliance reasons, accelerating development of self-developed alternatives.
This bidirectional decoupling may lead to efficiency losses in the short term. For developers, tools like Claude Code and GitHub Copilot have become part of daily workflows, and suddenly switching to less mature alternatives will impact productivity. However, in the long term, this may drive rapid maturation of China's AI tool ecosystem, similar to the leapfrog development achieved through local innovation in mobile payments, e-commerce platforms, and other domains over the past decade.
Regulatory Logic: Content Security Prioritized Over Technical Innovation
The core logic of China's current AI regulation wave is content security prioritization. The issuance of the Interim Measures for AI Anthropomorphic Interaction Services reflects regulatory agencies' forward-looking consideration of AI's potential social impacts. Unlike Western countries that focus more on AI bias, privacy, and competition issues, Chinese regulation emphasizes content orientation and social stability.
Anthropomorphic AI services have become a regulatory focus because these applications easily establish emotional connections with users, potentially influencing user values and behavioral patterns. Particularly among adolescent users, excessive reliance on AI companionship may affect development of normal social abilities. Additionally, user-customized agents without effective review may become channels for disseminating inappropriate content.
Notably, the new rules are relatively lenient toward enterprise-level and productivity AI tools, indicating regulatory agencies are seeking pragmatic paths between balancing innovation and control. For B2B applications, code assistants, data analysis tools, and similar scenarios, as long as they don't involve public emotional interaction and content dissemination, considerable development space remains.
Impact on Enterprises and Developers
For Chinese AI companies, this round of adjustments presents both challenges and opportunities. In the short term, disabling custom agent features will lead to user attrition and decreased engagement, particularly among user groups dependent on emotional companionship functions. However, in the long term, compliant operations help establish more sustainable business models, avoiding business fluctuations caused by regulatory uncertainty.
For enterprises and developers using AI tools, tool selection strategies require reassessment. Beyond functionality and performance, factors like compliance, data sovereignty, and supply chain stability have significantly increased in weight. Alibaba's promotion of Qoder to replace Claude Code reflects large enterprises constructing autonomous and controllable technology stacks.
For blockchain and digital asset industry participants, this trend equally warrants attention. Smart contract development, security audits, on-chain data analysis, and other processes increasingly rely on AI tools. Choosing compliant, trustworthy AI assistants and avoiding leakage of sensitive code and key information are important elements in ensuring asset security. Some institutional wallet and custody service providers have begun introducing stricter AI tool usage specifications in internal development processes, requiring code reviews to be conducted in isolated environments and prohibiting input of private key management-related code into public AI services.
Diverging Paths in Global AI Governance
China's AI regulatory measures exemplify the diversification of global AI governance. The EU has established a risk-based regulatory framework through the AI Act, the US relies more on industry self-regulation and ex-post accountability, while China has chosen a content-oriented pre-approval model. These three paths each have different emphases, reflecting different understandings of AI risks under different social systems and cultural backgrounds.
For AI companies operating across borders, this means developing differentiated product versions and compliance strategies for different markets. Anthropic's prohibition on Chinese entities using its models and OpenAI's restrictions on Chinese user access may be technically effective, but also objectively accelerate fragmentation of the global AI ecosystem.
From a technological development perspective, this fragmentation may lead to diversification of innovation paths. While Chinese AI companies are constrained by limited access to some frontier models, they are also exploring technical solutions better suited to local scenarios, such as federated learning with greater privacy protection emphasis and more efficient small model optimization. These explorations may provide new insights for global AI development in the future.
Enterprise Responses and Strategic Adjustments
The impact of these regulatory and security-driven changes extends beyond immediate compliance requirements. Chinese tech companies are accelerating investment in proprietary AI development capabilities. Alibaba's Qoder, ByteDance's internal tools, and Tencent's AI infrastructure represent strategic responses to reduce dependence on foreign AI services while building competitive advantages in the domestic market.
For multinational corporations operating in China, these developments necessitate careful navigation of complex compliance landscapes. Companies must balance global technology standards with local regulatory requirements, often requiring separate technology stacks, data handling procedures, and user interfaces for the Chinese market. This operational complexity adds costs but may be unavoidable for companies committed to the Chinese market.
The enterprise software and services sector faces particular challenges. Companies providing development tools, productivity software, and business intelligence solutions must ensure their AI-powered features comply with Chinese regulations while maintaining feature parity with global versions where possible. This may lead to diverging product roadmaps and increased localization efforts.
Implications for the Broader Tech Ecosystem
The tightening of AI tool controls in China has ripple effects across the global technology ecosystem. Open-source AI projects face questions about accessibility and contribution patterns when major markets implement usage restrictions. International collaboration on AI research and development becomes more complex when regulatory frameworks diverge significantly.
For developers and researchers, these changes create practical challenges in tool selection and workflow design. The fragmentation of the AI tools landscape means that skills and expertise developed with one set of tools may not transfer seamlessly across different markets. This could lead to the emergence of distinct developer communities with different toolchains and best practices.
Venture capital and investment patterns may also shift in response to these regulatory developments. Investors may favor companies with strong compliance frameworks and diversified technology dependencies. Startups building AI tools specifically designed for regulatory compliance or offering alternatives to restricted foreign services may attract increased attention.
Looking Ahead: Navigating Uncertainty
The current wave of AI tool controls in China represents a critical juncture in the evolution of global AI governance. As regulations continue to develop and enforcement mechanisms mature, companies and developers must remain adaptable and informed.
Several trends appear likely to continue: increased emphasis on data sovereignty and localization, growing investment in domestic AI capabilities, and continued divergence between Chinese and Western AI ecosystems. Companies succeeding in this environment will be those that can effectively navigate regulatory complexity while maintaining innovation momentum.
For the global AI community, these developments underscore the importance of understanding diverse regulatory approaches and their underlying rationales. While fragmentation poses challenges, it also creates opportunities for innovation in compliance technology, cross-border collaboration frameworks, and adaptive AI systems that can operate effectively across different regulatory regimes.
Conclusion: Balancing Security and Innovation
This round of AI tool controls by Chinese tech companies results from the intersection of regulatory policy, geopolitical factors, and technology security concerns. For industry participants, understanding regulatory logic, advancing compliance planning, and cultivating autonomous technical capabilities will be key to future competitiveness.
In the context of increasingly intense global AI competition, finding the balance between ensuring security and promoting innovation is a question all countries and enterprises must answer. China's exploration provides an observational sample, with effects and impacts that will gradually become apparent in coming years. For global AI practitioners, maintaining sensitivity to regulatory dynamics in different markets and building flexible response mechanisms will be necessary capabilities for navigating uncertainty.
Source: link