01
Models and computing power
Connecting self-reliant large models, APIs, tokens, computing resources and dedicated technical support.
Security Working Group
Cybersecurity Ecosystem
Leveraging the Alliance to connect professional users, real scenarios, in-depth data, capability evaluation and industry deployment, continuously moving agents from demonstration to production and jointly building a vendor-neutral, governable and continuously evolving self-reliant cybersecurity large-model ecosystem.
Continuous Loop
From one-off campaigns to a continuously running ecosystem loop
Building cybersecurity capability should not stop at a single study, validation or deployment. The working group turns scattered stage-based cooperation into a continuously running user network, task system, evaluation standards and industry relationships — a long-term mechanism that is cumulative and reusable.
01Model release
02Professional validation
03Scenario deployment
04Data accumulation
05Model enhancement
06Industry promotion
Key Resources
Six categories of key resources the Alliance connects
The Alliance’s value lies not in adding another organisation name but in long-term resource-organising capability: a professional-user entry + model release channels + a scenario-validation network + in-depth data sources + an industry results platform + a professional outreach matrix.
02
Academia and evaluation
Connecting university labs and expert teams to build benchmarks, dynamic ranges and real-task evaluation systems.
03
Data and tasks
Connecting security enterprises, competitions, ranges, open-source communities and offense-defense teams to form a continuous source of real tasks and in-depth data.
04
Professional talent
Connecting security researchers, offense-defense teams, university faculty and students, and professional developers into a stable testing, feedback and co-creation network.
05
Industry scenarios
Connecting security products, enterprise business, domestic software and key industries to bring model capabilities into real business scenarios.
06
Release and outreach
Connecting model release channels, industry media, professional communities and result platforms to promote and deploy new model capabilities professionally for the cybersecurity industry.
Priority Actions
Four priority actions
From professional-user operations to validation and release, forming a long-term mechanism that is participatory, cumulative and reusable.
01
Run the Cybersecurity Professionals Open Program on an ongoing basis
As the entry point for professional users, upgrade one-off benefit campaigns into a long-term user network: application review, benefit distribution, needs collection, tiered operations and accumulation of outstanding cases.
02
Launch the cybersecurity ecosystem co-building program
Open applications to security enterprises, labs, universities and professional researchers, offering closed-door exchange, benchmark/cyber-range testing, enterprise-grade model access and joint result releases.
03
Build an in-depth data and task system
Around real offense-defense, competition tasks, cyber mapping and dynamic ranges, build structured tasks for long-horizon work: task definition, environment tools, agent execution traces, decision processes, result adjudication and expert review — moving agents from demonstration to production.
04
Establish validation and application mechanisms for the cybersecurity industry
Around the needs of the cybersecurity industry, establish model validation and application mechanisms: trusted closed testing, professional validation and product-scenario deployment.
Ways to Participate
Multiple ways to co-build
No uniform data-delivery requirement and no mandatory delivery of raw data. Partners can take part through data, tasks, environments, experts, products, evaluation or scenarios.
- Authorized desensitized data
- Real offense-defense tasks
- Ranges and competitions
- Vulnerability environments
- Expert labeling and review
- New-model test feedback
- Security product integration
- Real business scenarios
- Domestic software and open-source security
Operating Mechanism
Organisation mechanism
Adopting “alliance coordination, model-vendor lead, academic guidance and partner co-building” — without building a complex new organisation, advancing through dedicated projects and joint actions.
Coordination and ecosystem hosting
Zhongguancun Independent Large Model Industry Alliance
Provides the organisational platform and industry credibility, coordinates resources across parties, reviews build directions and result releases, and takes on the platform and coordination role.
Model and technical support
Zhipu
Supports dedicated testing, release and capability iteration of new models; the ecosystem is equally open to different foundation-model vendors.
Academic guidance
Tsinghua University
Guides technical direction and evaluation methods, advances technical standards, research topics and talent development, and improves the professionalism and neutrality of results.
Industry research and ecosystem connection
Datainsecurity
Tracks industry trends and needs, connects security enterprises and professional partners, runs the Cybersecurity Professionals Open Program, and accumulates outstanding cases to promote outreach of results.
Platform operations, data and model training
CloudInfinite
Collects real offense-defense data and builds training and testing scenarios to continuously deepen and optimise the cybersecurity large model, focusing on scenario deployment for vulnerability discovery and penetration testing.
Ecosystem partners
Ecosystem co-building partners
Security enterprises, universities, labs and professional researchers contribute through data, tasks, experts, products or scenarios according to their strengths.
Responsible Governance
Governance boundaries
Public information is limited to ecosystem collaboration and governance principles; on data compliance and high-risk-capability governance, we proactively accept guidance from relevant authorities and experts.
- No disclosure of unauthorized sensitive materials or data; partners are not required to hand over raw data.
- No disclosure of high-risk capabilities or actionable attack details.
- Data compliance and public release of results require governance review and authorization confirmation.
Open Co-building
Different foundation-model vendors and technical ecosystems are welcome
The Cybersecurity Working Group upholds vendor neutrality and equal participation, and does not attribute participation eligibility, ecosystem capability or co-building results to any single technical path — jointly advancing China’s self-reliant cybersecurity large-model and agent systems.
- Vendor-neutral
- Equal participation
- Governable security
- Continuous evolution