CLAIR Acceptable Use Policy

Effective October 10, 2026 · Version 1.0

This policy is part of the CLAIR Terms of Use. It applies to everything you do with CLAIR, including with the AI models it runs on your computer. Breaking it can lead to suspension or revocation of your license under Section 7 of the Terms.

1. Use CLAIR lawfully

You must not use CLAIR to:

• break any law or regulation, or help anyone else do so;
• process personal data without a lawful basis or the required notices, consents or approvals;
• process data you are not allowed to use, including in breach of a data use agreement, confidentiality duty, research protocol, ethics or IRB approval, or your institution’s policies;
• infringe anyone’s intellectual property, privacy or other rights; or
• violate export control or sanctions laws.

2. Sensitive and regulated data

During normal use, CLAIR analyzes your data on your computer and does not send it to us. You decide what data goes into CLAIR, so you are responsible for handling it properly. If you load health information, student records, information about children, government identifiers, financial account data, or other regulated or confidential data:

• make sure you are authorized to use it on your device for your purpose;
• follow the safeguards your institution, contract or approval requires (for example, an encrypted, managed computer);
• do not send that data, or diagnostics files and screenshots that contain it, to us, to an outside AI service, or to anyone not authorized to receive it; and
• do not describe CLAIR as certified, compliant, approved or endorsed for any kind of data or by any regulator. If your institution has approved CLAIR for a specific use, you may say exactly that, and only that.

3. Decisions about people

CLAIR is intended for exploring and analyzing data, including for research. It has not been validated for making decisions about individual people. You must not use CLAIR’s AI outputs, predictions or statistics to decide, or to rank or screen people for, any of the following about a specific person:

• medical diagnosis, treatment or clinical care;
• admission, grading, discipline or assessment of a student;
• hiring, firing, pay or performance of an employee;
• credit, insurance, housing or access to essential services; or
• law enforcement, immigration or eligibility for public benefits.

You may use CLAIR to research these topics, for example to analyze outcomes across a past cohort, as long as the results are not used to make decisions about the individuals in the data. If you want to use CLAIR in a decision-making process, contact us first; that requires a separate written agreement.

4. Harmful uses

You must not use CLAIR or its AI models to:

• identify, track or profile people in ways they would not reasonably expect, including re-identifying people in de-identified data;
• discriminate against people on the basis of protected characteristics;
• create or spread content that is fraudulent, deceptive, defamatory or harassing, or that sexually exploits anyone, especially children;
• plan or promote violence, terrorism, self-harm, or the development of weapons;
• create malware or attack computer systems; or
• present AI-generated output as human-written, or as verified fact, when it has not been checked.

Studying these subjects is not prohibited. Lawful, authorized research that analyzes data about them (for example, research on suicide prevention, discrimination or cybersecurity incidents) is allowed, as long as it does not itself facilitate the harm and follows the stricter rules of the AI model you use (Section 5).

5. AI model licenses

Each AI model CLAIR downloads is licensed to you by its creator, not by us. Before a download starts, CLAIR names the exact model and its license. By downloading and using a model in CLAIR, you agree to follow its license and acceptable use policy. The models CLAIR currently offers come from these families:

Model family	Creator	License you must follow
Llama 3.1, 3.2 and 3.3	Meta	The Llama Community License for that version, and the Llama Acceptable Use Policy
Gemma 2 and Gemma 3	Google	Gemma Terms of Use and Gemma Prohibited Use Policy
Gemma 4	Google	Apache License 2.0
Qwen 2.5, Qwen 3, Qwen 3.5 and Qwen 3.6	Alibaba Cloud	Apache License 2.0 for the sizes CLAIR offers
Phi-3 and Phi-4	Microsoft	MIT License
Mistral and Mistral NeMo	Mistral AI	Apache License 2.0
DeepSeek-R1	DeepSeek	MIT License for DeepSeek’s own work. The 7B, 14B and 32B sizes are built on Qwen (Apache License 2.0) and the 70B size on Llama 3.3 (Llama license above)

If a model’s license or acceptable use policy prohibits something this policy allows, the stricter rule applies when you use that model.

6. Respect our service

You must not:

• bypass, disable or tamper with license checks or usage limits, or fake or change your device’s hardware identity;
• create multiple or fake accounts, use disposable email addresses, or sign up again after your license was revoked, to get around limits or a revocation;
• share, pool, sell or publish accounts, license keys or sign-in codes;
• probe, scan, overload or attack our servers, except as Section 7 allows; or
• use automated means to access our servers other than through the CLAIR app.

Using an email alias you own is fine. If you lose access to your account or need several accounts for an organization, contact us.

7. Reporting misuse and security problems

Tell us about misuse or security problems at [email protected]. We cannot see what you do in CLAIR, so we act on reports and on what we observe at our license servers.

Good-faith security research. You may test the CLAIR app on your own computer to look for security problems. Do not test our servers or anyone else’s account without our written permission. If you find a problem, report it to [email protected], give us 90 days to fix it before you publish details, and do not access, keep or share other people’s data. If you come across someone else’s data by accident, stop, tell us, and delete it. We will not pursue or support legal action against you for research that follows these rules. We cannot make that promise for other companies or for law enforcement.