One-sentence summary
The AI Safety Evaluation Framework for Industry and Information Technology (2026 Edition) was officially released on October 15, 2026 at the 2026 AI Industry Conference in Jinan; guided by the Cybersecurity Administration of the MIIT and compiled by the China AI Industry Development Alliance (AIIA) with relevant units, it defines evaluation content, methods and implementation requirements across general security, application security and business security.
Standard Overview
| Item | Detail |
|---|---|
| Name | AI Safety Evaluation Framework for Industry and Information Technology (2026 Edition) |
| Issuer | Guided by the Cybersecurity Administration of the Ministry of Industry and Information Technology (MIIT); compiled by the China AI Industry Development Alliance (AIIA) with relevant units |
| Status | Officially released October 15, 2026 at the opening of the 2026 AI Industry Conference |
| Location | Yellow River International Convention and Exhibition Center, Jinan, Shandong |
| Dimensions | General security, application security, business security |
| Content | Defines evaluation content, methods and implementation requirements |
Scope
- Developers, operators and third-party evaluation bodies of AI products and services in the industry and information technology field.
- Organizations that need AI security risk identification, protection-capability assessment and safety evaluation.
- Provides guidance for scientifically standardizing AI safety evaluation, covering manufacturing, telecommunications and related information-communication scenarios.
Key Provisions
| Dimension | Content | Affected Parties |
|---|---|---|
| General security | General safety evaluation content for AI systems | AI product and service developers, operators |
| Application security | Safety evaluation content for real AI application scenarios | Application developers, deployment operators |
| Business security | Safety evaluation content for business operations | Business operators, service providers |
| Evaluation methods | Defines evaluation methods and operating requirements | Evaluation bodies, developers, operators |
| Implementation | Defines organization, process and assurance requirements | Evaluation bodies, evaluated organizations |
Compliance Advice
- Organizations running AI businesses in the industry and IT field should build internal safety evaluation processes against the framework's three security dimensions.
- Developers and operators should regularly assess security risks and protection capabilities using the evaluation content and methods defined by the framework.
- Third-party evaluation bodies can use the framework as guidance for AI safety evaluation services.
- Organizations should track framework updates and supporting details, and use the latest version as the evaluation basis.
Official Sources
- Legal Daily (source: New Yellow River Client): AI Safety Evaluation Framework for Industry and Information Technology (2026 Edition) officially releasedlink