Core Guiding Principles for AI Development at Alibaba
People-Centricity
Inclusion and Integrity
Privacy Protection
Safety and Reliability
Trustworthiness and Controllability
Open Co-Governance
Model Governance
At Alibaba, we embed model governance—guided by our six principles for AI development—throughout the entire model lifecycle—from initial design and research through development, training, deployment, and real-world application—ensuring responsible AI is built in by design, not added on as an afterthought.
Safety & Alignment
Algorithm Design Phase
Model Training Phase
Safety Evaluation Phase
Content-Generation Phase
Use Guidance
AI Model Safety & Responsibility
Prohibited Uses of Qwen
Qwen is intended for lawful, constructive purposes and must not be used to cause harm. It cannot be used to exploit or endanger people, including children, or to promote violence, hate, discrimination, or abuse. Nor may it be used to invade privacy, engage in fraud or deception, spread misinformation, or bypass security safeguards. In short, Qwen is designed to support responsible and ethical use, and it must not be used in ways that undermine safety, rights, or public trust.
Specifically, we prohibit Qwen products and service from being used for:
Harm to minors
Creating, distributing, or promoting child sexual abuse material (CSAM) or content that exploits or harms minors.
Privacy violations
Compromising others’ privacy, including unauthorized data collection, surveillance, or biometric identification (e.g., facial recognition) without explicit consent. Uploading content containing personal data of minors without explicit consent from a legal guardian is prohibited.
Harmful content
Promoting violence, terrorism, hateful behavior, self-harm, or graphic violence.
Fraud and misrepresentation
Engaging in fraudulent activities, impersonation, or spreading disinformation/misinformation.
Exploitation
Facilitating human trafficking, labor exploitation, or abusive practices.
Emotion inference restrictions
Inferring the emotions of a natural person in the areas of workplace and education institutions, unless such use is permitted by applicable laws and regulations.
Recommendations for High-Risk Situations
Qwen is designed to support people’s work, not replace professional judgment in sensitive or high-impact situations. Where decisions carry serious consequences, human expertise and oversight remain essential. Qwen is intended to be used in ways that respect legal boundaries, public safety, and the stability of critical systems, with responsibility always resting with the people and organizations who deploy it. For high-risk decisions (e.g., healthcare, legal, financial), a qualified human professional must review AI outputs before finalizing or disseminating them.
Specifically, we consider the following as high-risk situations:
Healthcare
Do not use our services for medical diagnosis, treatment recommendations, or patient care without professional oversight by licensed professionals.
Legal and financial advice
Avoid providing tailored legal, tax, financial, or insurance advice without review by qualified professionals.
Critical infrastructure
Do not use our services to interfere with critical systems (e.g., power grids, healthcare databases, voting machines).
Law enforcement
Avoid using our services for official law enforcement, predictive policing, or surveillance unless explicitly permitted.
Protections for Minors
Qwen may not be used to target, groom, or exploit minors. It is strictly forbidden to generate content that sexualizes minors. Content moderation systems detect and block unsafe prompts and outputs, including harmful or inappropriate content involving minors. Photorealistic depictions of children are restricted where they may enable misuse or raise concerns around consent or legality, reflecting a zero-tolerance approach to exploitative content.
We use AI to support positive outcomes for minors, including applications for children with autism that generate stories and interactive content to enhance learning, accessibility, and well-being.
How Our Policy Is Enforced
Qwen model rules are enforced through a combination of automated checks, human review, and reports from users. Safeguards are regularly updated to reduce harmful or unsafe outputs. If rules are broken, the Qwen model may investigate, ask for corrections, issue warnings, remove content, limit accounts, or take legal action. For serious or repeated violations, access to Qwen model services may be suspended or ended. Users are encouraged to report any suspected violations through the provided channels.
