World Digital Technology Academy

PROGRAM & INITIATIVES

Assessment and Certification

Artificial Intelligence Safety, Trust & Responsibility

PROGRAM & INITIATIVES

AI STR Certification

Artificial Intelligence Safety, Trust & Responsibility

Artificial Intelligence Safety, Trust, and Responsibility (AI STR) Certification is a global program created by the World Digital Technology Academy (WDTA) based on the WDTA AI STR international standards. It provides a clear framework for ensuring that AI systems are secure, trustworthy, and responsibly governed.

The certification assesses both technical safeguards and ethical and legal compliance, helping organizations manage risks such as data protection, fairness, and broader societal impact. Achieving AI STR Certification demonstrates a commitment to responsible AI and strengthens trust with users, partners, and regulators. The program supports the global development of AI that is safe, transparent, and aligned with public interest.

AI Application

Evaluating and validating the performance of generative artificial intelligence (GenAI) applications in terms of safety, trust, and accountability

Large Language Model

Establish rigorous security standards and evaluation procedures to ensure large language models resist adversarial attacks, reduce risks, and promote responsible use.

AI Agent

The types of agent risks were further categorized, and testing methods such as model inspection, network communication analysis, and tool fuzz testing were improved and innovatively proposed.

Self-Assessment

Third-Party Audit

PROGRAM & INITIATIVES

Generative AI Application Security Certification

The Generative AI Application Security Certification is conducted based on the WDTA AI STR-01 standard.

The Generative AI Application Security Certification is conducted based on the WDTA AI STR-01 standard. It aims to assess and validate the performance of generative artificial intelligence (GenAI) applications in terms of security, trustworthiness, and responsibility. The certification covers multiple layers of testing and validation standards, including base model selection, embeddings and vector databases, prompt execution, agent behavior, fine-tuning, large model response processing testing, and runtime security of AI applications.

This certification is designed to ensure the security and reliability of AI applications throughout their lifecycle. It provides developers and organizations with a set of standards and guidelines to enhance the security of AI applications, mitigate potential security risks, improve overall quality, and promote the responsible development and deployment of AI technology.

Basic model selection

Embedding and Vector Database

Quick execution

Proxy behavior

The runtime security of artificial intelligence applications.

Large scale model response processing testing

fine-tuning

PROGRAM & INITIATIVES

Certification Basis

WDTA AI STR-01 Generative AI Application Security Testing and Validation Standard

Generative AI Application Security Testing and Validation Standard

Click to download

Target

Applications built on generative artificial intelligence (from LLM to GenAI)

This certification addresses key areas:

Base Model Selection

Prompt and Knowledge Retrieval (RAG)

Embedding and Vector Database

Prompt Execution/Inference

Agentic Behaviors

Fine-Tuning

Response Handling

PROGRAM & INITIATIVES

Large Language Model Security Certification

Based on the WDTA AI STR-02 Large Language Model Security Testing Method

The Large Language Model Security Certification is conducted based on the WDTA AI STR-02 standard. It aims to ensure that large language models can withstand various adversarial attacks, reduce potential risks, and promote their responsible application in real-world scenarios by establishing strict security testing standards and evaluation procedures. The certification includes comprehensive evaluations of models facing various attack types during pre-training, fine-tuning, and inference stages (such as data poisoning, backdoor attacks, instruction hijacking, and prompt obfuscation).

By setting clear evaluation metrics such as attack success rate (R) and refusal rate (D), the certification provides organizations with a scientific and systematic security detection method, helping them identify and fix potential vulnerabilities in the models, thereby enhancing the overall security and reliability of the models.

Additionally, the evaluation emphasizes protective measures for large language models in terms of data privacy, model integrity, and contextual applicability, ensuring that they can produce safe and reliable results when faced with adversarial examples.

PROGRAM & INITIATIVES

Certification Basis

WDTA AI STR-02 Large Language Model Security Testing Method

Target

Large Language Models

Large Language Model Security Testing Method

Click to download

PROGRAM & INITIATIVES

Value of LLM Security Certification

Enhancing Security and Reliability

Through rigorous security testing and evaluation, enterprises can identify and repair potential vulnerabilities in large language models, significantly improving the overall security and reliability of the models and reducing the risk of adversarial attacks.

Reducing Operational Risks

Through systematic security detection and protective measures, enterprises can effectively reduce operational risks caused by model security issues, avoiding potential economic losses and brand damage.

Promoting Responsible Use

Helps organizations consider social impact and ethical issues more thoroughly when designing and deploying AI systems, thereby promoting the responsible use of AI technology.

Cart