Acceptable Use Policy

1.Don't use the service to

• break the law, or help anyone else break it;
• create, distribute or solicit child sexual abuse material, or sexualise minors in any way. We report it to the authorities (in Canada, Cybertip.ca; in the United States, NCMEC) and close the account;
• develop or use weapons capable of mass casualties (biological, chemical, nuclear or radiological), or plan, promote or incite violence or terrorism;
• create malware, or attack, gain unauthorised access to, or disrupt any system, network or service;
• harass, threaten, bully or defame people, or promote hatred against people based on protected characteristics;
• defraud or deceive people: phishing, scams, spam, fake reviews, impersonation, or passing AI output off as a human's where the law or context requires that it be disclosed;
• violate privacy: covert surveillance or tracking, unlawful collection or processing of personal information, or inferring sensitive characteristics (such as health, sexuality or political views) about people without a lawful basis;
• make automated decisions with legal or similarly significant effects on people (credit, employment, housing, insurance, education, legal status, medical care) without appropriate human review and the notices the law requires;
• infringe intellectual property or other rights, including by violating a model's licence;
• give individualised medical, legal or financial advice as if it came from a licensed professional, without one involved.

2.Don't abuse the service

• Don't create multiple accounts, or rotate networks or identities, to get around signup limits, rate limits, budgets, suspensions or payment requirements.
• Don't share, sell or publish API keys, or let others use your account except through your own product.
• Don't probe, scan or load-test our systems, or try to reach another customer's data or another region's systems. Security research is welcome under our vulnerability disclosure policy.
• Don't interfere with the service, other customers, or our upstream providers, or try to extract our credentials or theirs.
• Don't use the service to send content you don't have the right to send, or to send it somewhere you have promised it won't go.

3.Model licences

Several licences go beyond permissive open-source terms. For example, Meta's Llama 4 Community License incorporates Meta's acceptable use policy, requires “Built with Llama” attribution in products that use it, and requires a separate licence from Meta for services above 700 million monthly active users. Other custom licences can require attribution, restrict certain uses or apply thresholds based on your size or revenue. Read the licence before you build on a model; it is linked from the model's page and from the table below.

Licence names come from each publisher's model card and can change with new releases; the publisher's licence text is what applies. Output restrictions in a licence (for example, on using output to train other models) apply to you too.

4.Your users and your product

If you build a product on ScaledThought, you are responsible for how your users use it. Your own terms must be at least as protective as this policy, and you must act on misuse you become aware of.

5.Enforcement and reporting

We don't log or read your content, so we don't screen it in advance. We act on reports, on legal requests and on signals in usage metadata. If we believe this policy has been broken, we may warn you, revoke keys, remove credits obtained through abuse, or suspend or close the account, depending on the severity. Where content suggests a risk to someone's life, or child sexual abuse, we may report it to the authorities.

Report misuse to support@scaledthought.com, and security issues to security@scaledthought.com.