New policy takes effect, AI API calls enter compliance era
In early November 2026, the Ministry of Industry and Information Technology and the Cyberspace Administration of China jointly issued the "Management Measures for API Calls of Generative AI Services," requiring all enterprises that use APIs to call third-party large models to route through a compliant proxy platform with content review and security audit capabilities, and to meet traceability and auditability standards for call logs and sensitive content interception. This policy directly impacted the previous crude approach of enterprises directly connecting to model vendor APIs, and also spawned a new infrastructure category – the AI API relay station.
After the new regulations took effect, enterprises' demand for "legally compliant large model calls" surged. Many enterprises found that building their own content review and logging systems was costly and difficult to handle interface differences across multiple model vendors. The market urgently needed a unified gateway that could meet compliance requirements while reducing total cost of ownership.
AI hub launches AI large model proxy service: unified access, compliant relay, token wholesale
On November 12, AI hub officially launched the "AI大模型代理" service, providing a one-stop large model API calling platform for enterprises and developers. This service is essentially an AI channel proxy: by connecting to multiple domestic mainstream model vendors (such as Qianwen, Wenxin, Zhipu, etc.), it performs request forwarding, content security review, usage statistics, and billing optimization in the middle layer. Enterprises need only a single integration to call multiple models, while automatically meeting the new regulations' requirements for content auditing and log retention.
As an AI API relay station, the service introduces a token wholesale mechanism for cost. AI hub negotiates tiered API pricing with upstream model vendors and aggregates call volumes from multiple customers to obtain lower unit prices, then offers them to downstream enterprises in a "token wholesale" model. Internal calculations by AI hub show that compared to enterprises purchasing directly, using this proxy service can reduce token usage costs by approximately 30% to 50%, making it one of the industry's recognized "cheapest token" acquisition channels. Additionally, for latency-sensitive scenarios, AI hub has deployed edge nodes in major regions, supporting request caching and dynamic routing to maintain compliance while approaching direct connection performance.
Service architecture and core capabilities: more than a proxy, a compliance foundation
The AI large model proxy service consists of three core modules:
- Unified gateway: Provides standard RESTful and gRPC interfaces, automatically compatible with request format conversion across model vendors; enterprises can switch models without modifying existing code.
- Intelligent audit engine: Deployed in the forwarding chain, performs real-time multi-dimensional content security detection on all inputs and outputs (including sensitive information such as pornography, violence, rumors), and generates non-tamperable call logs to meet regulatory requirements.
- Billing and optimization engine: Supports multiple billing modes such as per token, per call, and per time period; built-in caching strategy automatically returns cached results for repeated requests with the same prompt, further reducing actual token consumption.
Notably, AI hub has deeply integrated the token代理 service with its original model management platform, allowing enterprises to view call volumes, latency, error rates, and compliance interception records for all models from the same console. This architecture not only solves the compliance pain points under the new regulations but also provides a foundation for enterprises to carry out multi-model hybrid scheduling and cost reduction and efficiency improvement.
Investment and industry response: compliance proxy becomes capital's new favorite
The policy implementation and actions by AI hub and other platforms have also attracted capital market attention. Since November, multiple AI infrastructure startups have announced new funding rounds, with "AI API relay station" type projects particularly sought after. A startup named "LinkToken" disclosed that its Series A funding exceeded 100 million yuan, with investors explicitly valuing its first-mover advantage in "token wholesale" and compliance auditing. Industry analysts pointed out that with regulatory tightening, AI large model proxy services like AI hub's will become standard gateways for enterprises deploying large model applications, and the market size is expected to exceed tens of billions of yuan next year.
At the same time, some large model vendors have begun to adjust strategies, actively cooperating with proxy platforms to broaden application scenarios. Some vendors said that the proxy model reduces their support pressure when dealing directly with numerous small and medium-sized customers, while also maintaining reasonable profits through aggregated bargaining by proxies. Overall, the AI channel proxy is transforming from the old impression of "middleman making a spread" to a new role of "compliance and cost optimization foundation."
In this wave of policy and market resonance in November 2026, AI hub took the lead by deploying the AI large model proxy service, responding to enterprises' urgent compliance needs while lowering the barrier to AI application through innovative models like token wholesale. In the future, as more industry scenarios connect, such proxy platforms may become a key link in driving the universal adoption of large models.