China's CNCERT flags rogue AI skill packages for crypto mining, guardrail bypass
Synopsis
China's CNCERT has publicly flagged a grey market of AI 'skills' packages that claim to bypass model safety guardrails and unlock banned cryptocurrency-mining functions — a rare regulatory strike directly at the AI extension layer, not just the underlying models.
Key Takeaways
China's CNCERT issued a formal warning on 10 June 2026 against unregulated third-party AI skills packages via its official WeChat account.
Some skills are marketed to circumvent built-in AI model safety guardrails , enabling generation of prohibited content.
Other packages reportedly provide access to cryptocurrency-mining functions, which have been banned in mainland China since 2021 .
CNCERT warned that using such tools risks privacy breaches , account suspensions , data leaks , money-laundering exposure and legal liability.
The agency recommends sourcing AI skills only from official channels and applying the least-privilege principle when granting permissions.
The alert extends China's AI regulatory scrutiny from foundational models to the application and extension layer for the first time.
China's top cybersecurity authority has issued a formal warning against unauthorised third-party AI 'skills' packages being marketed to bypass model safety guardrails and enable cryptocurrency mining — activities that remain banned on the Chinese mainland. The National Computer Network Emergency Response Coordination Centre (CNCERT) published the alert on Tuesday, 10 June 2026, via its official WeChat account, flagging a rapidly expanding grey market for unregulated AI extensions.
What Are AI Skills and Why Do They Matter
In the AI ecosystem, skills function as plug-ins or specialised code packages that extend the capabilities of AI agents and models. Much like smartphone apps, they can connect AI systems to external databases, automate workflows and integrate with third-party software or online services, enabling tasks well beyond standard text generation.
The category has grown swiftly alongside the proliferation of large language models and autonomous AI agents, creating a parallel distribution layer that regulators are only beginning to scrutinise.
The Threat Landscape CNCERT Identified
CNCERT said some skills packages are being marketed explicitly as tools for circumventing built-in restrictions in AI models, allowing users to generate otherwise prohibited content. Others reportedly provide access to cryptocurrency-mining functions — a category of activity that has been prohibited in mainland China since 2021.
The agency warned that deploying such tools could expose users to privacy breaches, account suspensions, data leaks, money-laundering risks and potential legal consequences under existing Chinese law.
Why It Matters
The alert signals that China's regulatory apparatus is extending its oversight from foundational AI models — already subject to the country's Generative AI Interim Measures — down to the application and extension layer. This mirrors global concerns about agentic AI systems that can autonomously interact with external services, raising accountability gaps that existing frameworks were not designed to address.
The warning also arrives as domestic AI adoption accelerates, with enterprises and individual developers increasingly building on top of models from companies such as Baidu, Alibaba and DeepSeek, creating fertile ground for third-party extension marketplaces.
CNCERT's Recommended Safeguards
CNCERT advised users to obtain AI skills exclusively through official, verified channels and to apply the principle of least privilege when granting permissions to any extension. Users should also promptly revoke unnecessary access to sensitive data once a task is complete, the agency said.
The guidance draws a clear line between the legitimate AI skills ecosystem — which the agency acknowledged includes many compliant offerings — and the grey-market tools it is targeting.
What's Next
With CNCERT's warning now public, platform operators and app stores hosting AI skill repositories face implicit pressure to tighten vetting procedures. Analysts expect follow-on enforcement guidance from the Cyberspace Administration of China (CAC) as the country moves to close regulatory gaps in the agentic AI stack before they become systemic vulnerabilities.
Point of View
Not a workaround. Globally, this mirrors debates in the EU and US about liability for agentic AI plugins, but China's approach — a public warning with implicit platform liability — is faster and less process-heavy than Western rulemaking. What mainstream coverage underplays is the crypto-mining vector: embedding mining code inside an AI skills package is an elegant evasion of China's blanket crypto ban, and the fact that it has surfaced in a cybersecurity bulletin suggests it is already widespread enough to warrant official attention.
NationPress
25 Jul 2026
Frequently Asked Questions
What did China's CNCERT warn about regarding AI skills?
China's CNCERT warned on 10 June 2026 that certain third-party AI skills packages are being marketed to bypass model safety guardrails and enable banned cryptocurrency-mining functions. The agency said these tools expose users to data leaks , money-laundering risks, account suspensions and potential legal consequences.
What are AI skills or AI skill packages?
AI skills are plug-ins or specialised code packages that extend the capabilities of AI agents and models, similar to smartphone apps. They can connect AI systems to external databases, automate workflows and integrate with third-party software, enabling tasks beyond basic text generation.
Why is cryptocurrency mining banned in China?
Mainland China banned cryptocurrency mining and trading in 2021 , citing financial stability risks and energy consumption concerns. Any software — including AI skill packages — that enables mining activity therefore falls outside the law.
How can users protect themselves from rogue AI skill packages?
CNCERT recommends obtaining AI skills only from official, verified channels and applying the principle of least privilege when granting permissions. Users should also promptly revoke any unnecessary access to sensitive data after completing a task.
What does CNCERT's warning mean for China's AI regulation?
The alert signals that China's regulatory reach is extending beyond foundational AI models to the application and extension layer. It is likely to increase pressure on platform operators hosting AI skill repositories and may prompt follow-on enforcement rules from the Cyberspace Administration of China (CAC) .
