router.africa Model Terms
Effective date: 19 September 2026
1. Why this document exists
1.1 The models on router.africa are built and run by other companies (“Upstream Providers”). Each sets its own contractual terms and usage rules for its models (“Upstream Terms”). This document explains how those rules reach you and forms part of our Terms of Service.
1.2 Whenever you call a model you agree to follow (a) these Model Terms, (b) the Upstream Terms for that model, and (c) any extra conditions shown on the model’s listing in the dashboard. You must not use the Platform in a way that would put you or us in breach of an Upstream Provider’s rules.
1.3 If a model’s listing or Upstream Terms are stricter than our Terms of Service on some point, the stricter rule governs use of that model.
2. Where to find Upstream Terms
Each model’s listing in the dashboard links to its Upstream Provider’s terms and usage rules, and shows what we know about that provider’s data retention and training practices. This information reflects our best current understanding and does not replace reading the owner’s own terms, which they may change at any time. Where we have a special arrangement with an Upstream Provider, the model’s listing will say so.
3. Your duty to follow Upstream Terms
3.1 You are responsible for reading and following the Upstream Terms for every model you use. These commonly cover: banned content and activities, sensitive or regulated uses, whether responses may be used to train other models, competing with the owner, age and disclosure rules, country and sanctions limits, and sharing or reselling access.
3.2 If an Upstream Provider requires your end users to accept its terms, or requires you to pass along certain notices or restrictions, you must do so.
3.3 If we tell you an Upstream Provider objects to your usage, or an end user’s usage, you must stop or fix it promptly. We may pause routing to that provider, or pause your account, until it’s resolved.
4. Limited models, uses and locations
4.1 Some models, and some uses of them, are limited by their owners, by law or by our own risk review. Limits may depend on your country, sector, purpose or volume. We may show the limit on the model’s listing, return an error or decline the request.
4.2 Upstream Providers can tighten or relax limits whenever they like, sometimes without telling us first. We’ll pass changes to you as soon as we reasonably can, but access may be blocked before you hear from us.
4.3 You must not use tools, proxies or false information to hide where you are, who you are or what you’re doing in order to get past a limit.
4.4 You must not use a model if you’re located or ordinarily resident somewhere the Upstream Provider or the law prohibits access, or if you are a sanctioned person or entity.
5. Using responses to train other models
5.1 Some Upstream Providers forbid using a model’s responses to train, fine-tune or teach another model, or to build a rival product. A model’s listing will say whether that is allowed. You must follow the rule that applies to each model whose responses you use.
5.2 If your use case depends on this permission, tell us at legal@router.africa and choose only models whose listing allows it.
6. How Upstream Providers treat data
6.1 To serve your request, your Content goes to the Upstream Provider. Some keep prompts and responses for safety monitoring or legal reasons, and some may use them to improve their own models. Practice varies by provider, by model, by endpoint and sometimes by pricing tier.
6.2 Where routing controls are available, they are described in our documentation. We can’t promise that any request will be handled only by providers with a particular data policy.
6.3 You decide whether a provider’s data practices are acceptable for the personal or confidential data you send, and you remain responsible for your own data-protection duties, including cross-border transfers. See our Privacy Policy.
7. What models can and can’t do
7.1 Models work on probabilities. They can produce wrong, incomplete, dated, biased, offensive or unsafe responses, can invent facts or sources, and may answer the same prompt differently each time.
7.2 Models differ in ability, context size, supported settings and built-in safety behaviour. A model may ignore a setting it doesn’t support.
7.3 Some providers run compressed or lower-precision versions of a model, which can change quality. Where we know the precision level we show it on the listing.
7.4 Preview, beta, experimental and free models can be less dependable, may change or vanish without notice, may follow different data rules and carry no availability commitment.
7.5 Responses are not medical, legal, financial, tax, engineering or other professional advice. Don’t let a response alone decide something with legal or similarly serious consequences for a person; put a qualified human in the loop.
7.6 Never rely on a model as the only safeguard where a mistake could cause death, injury or serious harm.
8. Being open and responsible with your users
8.1 If your application shows AI-generated material, tell your users so, in a way that suits your context and any law that applies.
8.2 Apply content controls, rate limits, abuse monitoring and human escalation that fit your application’s level of risk. Don’t strip out safety notices or provenance labels that an Upstream Provider requires you to keep.
8.3 Review responses before publishing or acting on them, and check that your use respects others’ intellectual property, privacy and rights.
9. Availability, pricing and retirement of models
9.1 Prices, context limits, supported settings and availability follow what the Upstream Provider offers and can change. The price shown on a model’s listing when you send a request is the price for that request.
9.2 Upstream Providers retire models. We’ll try to relay announced retirements and suggest alternatives, but we can’t promise that a model will stay available or behave the same over time.
9.3 We may add, remove or re-rank Upstream Providers for any model to protect availability, cost or compliance. We don’t promise a specific provider will handle a specific request unless you’ve limited routing to that provider (and then only while it’s up).
10. Rights in responses
10.1 Whether a response is protected by copyright or any other right, and who holds it, depends on the Upstream Terms and applicable law, and may be limited or uncertain. As between you and us, we claim no ownership of responses, and we don’t promise that they’re free of third-party rights.
10.2 You answer for any claim that your Prompts, or your use of responses, infringes or misuses someone else’s rights.
11. No endorsement; where our responsibility stops
11.1 Listing a model or provider isn’t a recommendation. We don’t build, control or audit models or providers and don’t vouch for their performance, safety or compliance.
11.2 To the fullest extent the law allows, and subject to our Terms of Service, we aren’t responsible for an Upstream Provider’s outages, changes, restrictions, data practices or responses, or for loss that results from one suspending or limiting your access.
12. Updates
We may update these Model Terms and the Upstream Terms information on model listings at any time. We’ll post the new version and give notice of material changes as our Terms of Service describe. Using a model after a change means you accept it.
13. Contact
Model Terms questions or suspected Upstream Terms issues: legal@router.africa