Loading the catalogue…
Loading the catalogue…
BAAI (Beijing Academy of Artificial Intelligence / Zhiyuan Institute) is a Chinese government-funded nonprofit research institute established in 2018 under the guidance of China's Ministry of Science and Technology and the Beijing municipal government. It is one of China's most influential AI labs, widely known for the BGE embedding model family and the WuDao large model series, and distributes models under a mixed open-source policy. Critically for EU regulated customers, BAAI was added to the US Bureau of Industry and Security Entity List in March 2025 on grounds of alleged AI development in support of China's military modernisation; while this carries no direct EU legal bar on use, it signals elevated dual-use risk, documented ties to military-adjacent institutions, and reputational exposure for organisations subject to export-control compliance obligations. BAAI publishes no GDPR-oriented privacy policy, no EU AI Act compliance statement, and no ISO 27001 or SOC 2 certifications, making independent verification of data-handling and security practices impossible for EU enterprise customers.
BAAI was added to the US Bureau of Industry and Security Entity List on 25 March 2025 (Federal Register 2025-05427) for allegedly developing large AI models and advanced computing chips for China's military modernisation. A licence is required for all US-origin items, with a presumption of denial. This is the most significant risk signal for EU regulated customers.
BAAI is funded primarily by the Beijing municipal government and was established under guidance of China's Ministry of Science and Technology. This creates structural government dependency and alignment with Chinese state AI priorities, including potential military-civil fusion obligations under Chinese law.
BAAI researchers co-authored a paper with the director of Peng Cheng Lab — a military-affiliated institution independently listed on the US Entity List. The extent of ongoing collaboration is unverified.
BAAI operates exclusively under Chinese law (PIPL, DSL, Cybersecurity Law, National Intelligence Law). China's National Intelligence Law (Art. 7) requires organisations to 'support, assist, and cooperate with state intelligence work' — a compelled-access risk analogous to CLOUD Act exposure, but under Chinese jurisdiction.
No GDPR-oriented privacy policy, DPA template, DPO contact, or EU-adequate data transfer mechanism has been published. China is not an EU-adequate jurisdiction. Processing of personal data involving BAAI as a data processor would lack a valid legal basis under GDPR.
No EU AI Act compliance statement, no GPAI Code of Practice signature, and no EU-format training data summary published. BAAI's most widely deployed models (BGE family) are already in scope of the Act's GPAI obligations as of August 2025 for new placements.
Stav AI Act assessment
Editorial assessment, not legal advice. Stav's risk ratings, scores, and verdicts are our own analysis of publicly available information and may be incomplete or out of date. Verify independently before making compliance or procurement decisions.
BAAI maintains one of the most active open-source AI profiles globally, with approximately 169 models and 120+ datasets on HuggingFace, all publicly accessible. BGE-large-en-v1.5 alone exceeds 14 million monthly downloads, demonstrating substantial real-world validation.
BAAI's BGE, Aquila, and FlagOpen models are released under permissive licences (MIT, Apache 2.0), enabling customers to inspect weights, reproduce results, and perform independent security evaluation.
BAAI led the release of the Beijing AI Principles (2019), one of the earliest national AI ethics frameworks, endorsed by Tsinghua, Peking University, and the Chinese Academy of Sciences — demonstrating early and sustained engagement with responsible AI governance.
BAAI hosted China's first AI safety international dialogue (March 2024) with participants including Nobel laureate Geoffrey Hinton, positioning it as a significant actor in global AI safety discourse.
BGE models have been adopted by Microsoft Azure, Amazon AWS, Alibaba Cloud, Tencent Cloud, Huawei Cloud, and integrated into LangChain, LlamaIndex, and HuggingFace Transformers — indicating broad third-party technical validation.
BAAI continues to operate and publish new models and research (including the WuJie series and new robotics work) following the US Entity List designation, with no evidence of operational contraction.
Extensive peer-reviewed publication record including CVPR 2023 and ICCV 2023 papers; active ArXiv preprint presence as recently as June 2025. Research outputs subject to international academic peer review.
Privacy policy review
Creator profile
Stav compliance has not yet scored Beijing Academy of Artificial Intelligence. Scores are published once the policy review and infrastructure assessment complete.
Findings
Citations gathered when the Compliance Curator last reviewed this creator’s public-facing documents. Grouped by source so the picture stays auditable.
“Founded on November 14, 2018 with backing from the Beijing Municipal Science and Technology Commission and the Zhongguancun Science Park Management Co...”
“As a collaborative hub, BAAI's founding members include leading AI companies, universities, ... Beijing Academy of Artificial Intelligence (BAAI...”
As a collaborative hub, BAAI's founding members include leading AI companies, universities, ... Beijing Academy of Artificial Intelligence (BAAI...
The ERC determined to add the Beijing Academy of Artificial Intelligence and Beijing Innovation Wisdom Technology Co., Ltd. to the Entity List, under ...
The Beijing-based agency said it was shocked by the decision as it was a “non-governmental, non-profit research institution” that has made its researc...
Founded on November 14, 2018 with backing from the Beijing Municipal Science and Technology Commission and the Zhongguancun Science Park Management Co...
Observers say that may have to do with BAAI’s work with a Chinese military-affiliated entity known as Peng Cheng Lab. In 2021, Huang Tiejun, BAAI’s de...
As a collaborative hub, BAAI's founding members include leading AI companies, universities, ... Beijing Academy of Artificial Intelligence (BAAI...
The ERC determined to add the Beijing Academy of Artificial Intelligence and Beijing Innovation Wisdom Technology Co., Ltd. to the Entity List, under ...
The Beijing-based agency said it was shocked by the decision as it was a “non-governmental, non-profit research institution” that has made its researc...
Founded on November 14, 2018 with backing from the Beijing Municipal Science and Technology Commission and the Zhongguancun Science Park Management Co...
Observers say that may have to do with BAAI’s work with a Chinese military-affiliated entity known as Peng Cheng Lab. In 2021, Huang Tiejun, BAAI’s de...
No security certifications (SOC 2, ISO 27001), bug bounty programme, or responsible disclosure policy identified. Security posture cannot be independently verified by EU enterprise customers.
European companies using US-origin technology (e.g. GPU compute, cloud infrastructure) to deploy, fine-tune, or re-train BAAI models may face export-control compliance obligations given BAAI's Entity List footnote 4 designation, which covers foreign-produced items subject to the EAR.
Published safeguards & certifications
“The ERC determined to add the Beijing Academy of Artificial Intelligence and Beijing Innovation Wisdom Technology Co., Ltd. to the Entity List, under ...”
“The Beijing-based agency said it was shocked by the decision as it was a “non-governmental, non-profit research institution” that has made its researc...”
“Observers say that may have to do with BAAI’s work with a Chinese military-affiliated entity known as Peng Cheng Lab. In 2021, Huang Tiejun, BAAI’s de...”
As classified under Regulation (EU) 2024/1689.
Provider of GPAI model (general-purpose).