What is an acceptable use policy?
Written by Brooke Ash, lawyer at For Founders Law.
Short answer
An acceptable use policy (AUP) is the set of rules for what users may not do with your product, such as unlawful activity, spam, security attacks, infringing content or, for AI products, generating harmful or deceptive output. It usually sits beside your terms of service and is made part of them, so a breach is a breach of contract that lets you suspend or close the account. To be enforceable, users must agree to it, it must be specific, and your terms must give you the right to act on it.
We can help at a fixed fee: Terms of Service, from $500. Book a free 15-minute consult
Most founders first meet an acceptable use policy as a link in someone else’s footer. You need your own once users can upload content, send messages, call your API or prompt your AI features, because those are the ways your product can be misused in your name.
What is an acceptable use policy?
An acceptable use policy, or AUP, is a list of things users may not do with your product. It covers conduct, not commercial terms: no illegal activity, no spam, no attacks on the service, no infringing or abusive content, and for AI products, no using the model to produce harmful, deceptive or high-risk output. The terms of service then say that breaking the AUP is a breach of contract and that you can suspend or end the account.
What’s the difference between an AUP and terms of service?
The terms of service are the whole contract. The AUP is one part of it, usually kept as a separate page so you can point to it and update it more easily.
| Terms of service | Acceptable use policy | |
|---|---|---|
| What it covers | The full relationship: fees, licence, IP, liability, termination, governing law | What users must not do with the product |
| Who it mainly protects | Both parties, through agreed terms | You, other users and third parties, by limiting misuse |
| How it’s agreed | Clickwrap or signed agreement | Incorporated by reference into the terms, so it’s accepted at the same time |
| How often it changes | Rarely, with notice | More often, as new misuse appears |
| What happens on breach | Remedies set out in the terms | Suspension or termination under the terms |
An AUP that isn’t tied to the terms is a set of house rules with no contractual force. Make sure the terms say users agree to it, link it from the acceptance screen, and say how changes are notified. See do I need terms of service? and are clickwrap agreements enforceable?
What should a SaaS acceptable use policy include?
The core list is similar for most products:
- Unlawful use, including fraud, and content that is defamatory, harassing, hateful or that sexually exploits anyone.
- Content that infringes copyright, trademarks or privacy rights, or that the user has no right to upload.
- Spam and unsolicited commercial messages sent through the product, which can also breach Canada’s anti-spam law.
- Security abuse: probing, scanning or testing vulnerabilities without permission, bypassing access controls, spreading malware, or interfering with the service.
- Overloading the service, scraping it, or exceeding rate limits.
- Reselling or sharing access beyond what the plan allows.
- Using the product to build a competing product, if that matters to you.
- Uploading personal information the user isn’t authorized to share, or sensitive information the product isn’t built to handle, such as health records in a tool that isn’t set up for them.
Write the list for your product. A generic AUP copied from a large platform will ban things your product can’t do and miss the misuse it actually invites.
What should an AI product’s acceptable use policy add?
AI features invite their own kinds of misuse. Common additions:
- Generating content that impersonates a real person or organization, or deepfakes of real people without consent.
- Using output for decisions about people’s employment, credit, housing, insurance or health without meaningful human review, unless the product is designed and contracted for that use.
- Presenting AI output as human-written where that would mislead, such as fake reviews or testimonials.
- Trying to extract training data, system prompts or model weights, or to bypass safety limits.
- Using the product to give professional advice, such as legal, medical or financial advice, to others as if a professional had checked it.
If you build on a third-party model, check the provider’s usage policy. Providers such as OpenAI publish usage policies and expect you to pass those restrictions on to your users, so your AUP should be at least as strict as theirs. See what foundation model API terms mean for your product and terms for AI products.
How do you enforce an acceptable use policy?
Through the suspension and termination rights in your terms. A workable approach:
- Reserve the right to remove content, suspend features or suspend the account immediately where there’s a security risk, legal risk or harm to others.
- For other breaches, give notice and a chance to fix the problem before suspending.
- Say what happens to fees: whether prepaid fees are refunded if you terminate for breach.
- Say what happens to the user’s data, including whether they can export it after termination.
- Keep a record of what you saw and why you acted.
Enterprise customers often negotiate this. Expect requests for notice before suspension except in emergencies, and suspension limited to the affected users or features. Settle your position in your SaaS agreement before the first enterprise deal.
If you review users’ content or activity to enforce the AUP, say so in your privacy policy. PIPEDA’s openness principle expects you to be clear about how you handle personal information, and users shouldn’t learn about content review from a suspension notice.
Are there Canadian rules that affect an AUP?
A few that come up often:
- Copyright notices. Canada uses a “notice and notice” system. A business that hosts content for users may receive a notice of claimed infringement under section 41.25 of the Copyright Act, and must forward it to the user and keep identifying records, generally for six months. Failing to do so exposes the business to statutory damages of $5,000 to $10,000 under section 41.26. Banning infringement in your AUP doesn’t replace that process.
- Anti-spam. If users can send email or messages through your product, ban messages that breach CASL and build in consent and unsubscribe tools where you can.
- Consumer terms. Ontario’s consumer protection rules for internet agreements apply to consumer users, so keep consumer and business versions of your terms separate where you can. See online terms and Ontario consumer protection.
What to do next
We draft terms of service with an acceptable use policy tailored to your product, and AI product terms that line up with your model provider’s usage policies, at fixed fees on our fees page.
This guide is general information about Ontario and Canadian law, not legal advice for your situation. To talk through your terms before launch, book a free 15-minute consult.
Sources
- Electronic Commerce Act, 2000 (Ontario), s. 19 (acceptance by clicking)
- Copyright Act, R.S.C. 1985, c. C-42, s. 41.25 (notice of claimed infringement, including to hosting providers)
- Copyright Act, s. 41.26 (obligations on receiving a notice; statutory damages of $5,000 to $10,000)
- Canada's Anti-Spam Legislation, S.C. 2010, c. 23
- PIPEDA, Schedule 1, Principle 4.8 (openness)
- OpenAI Usage Policies
Citation copied
How we can help
- SaaS or Master Services AgreementYour customer contract, ready for business customers. From $920
- AI Product TermsInputs, outputs and training use, in plain terms. From $760
Fixed fees. Government filing fees are extra.