Alex is Sprintlaw’s co-founder and principal lawyer. Alex previously worked at a top-tier firm as a lawyer specialising in technology and media contracts, and founded a digital agency which he sold in 2015.
Maya runs a London AI SaaS startup that helps HR teams shortlist internal training content. A large customer is ready to sign, but their procurement team has asked a difficult question: who is responsible if the AI output is wrong?
That one question can slow down a launch, a funding round or an enterprise sale. Your standard SaaS terms might cover payment, access and subscriptions, but AI products usually need extra wording around data inputs, outputs, model limits, acceptable use and liability.
The good news is that this doesn't have to be complicated. Well-drafted AI SaaS terms can set clear expectations, protect your business and give customers confidence that you've thought through the real risks.
Do AI SaaS Terms Need To Be Different From Standard SaaS Terms?
Yes, AI SaaS terms usually need extra clauses because your product doesn't just host software - it generates, predicts, recommends or classifies something for the customer.
Standard SaaS terms and conditions in the UK often focus on access to the platform, payment, uptime, support, account suspension and intellectual property. Those are still essential. But AI adds a different set of practical questions:
- What can the customer put into the system?
- Who owns or can use the AI output?
- Can the customer rely on the output without checking it?
- Can you use customer inputs to improve your model?
- What happens if the output is inaccurate, biased or incomplete?
- Who is responsible for complying with laws in the customer's industry?
If your terms don't answer these questions, you may end up negotiating them from scratch with every serious customer. Worse, the customer may assume you're taking responsibility for things you never intended to guarantee.
For example, if your product drafts marketing copy using AI, your terms should make clear that the customer is responsible for reviewing the output before publishing it. If they publish a misleading claim without checking it, you don't want that treated as your fault simply because your platform produced the first draft.
Your platform terms and conditions should match how your product actually works. A customer support chatbot, an AI accounting assistant and a legal document automation tool all need different risk settings.
What Should Your AI Software Terms Of Use Say About Inputs?
Your AI software terms of use should clearly say what customers can and can't upload, because the quality and legality of the input often affects the output.
Inputs are the prompts, documents, data, images, audio, code or other material your customer puts into your platform. For AI SaaS businesses, inputs are often where privacy, confidentiality and intellectual property risks begin.
Your terms should require customers to confirm that they have the right to provide any input they upload. This is especially important if they might upload personal data, confidential business information, third-party content or regulated data.
Common Input Rules To Include
- Permission: The customer must have all rights, licences and consents needed to upload the input.
- Accuracy: The customer is responsible for making sure inputs are accurate, lawful and complete.
- Prohibited material: The customer can't upload illegal, harmful, infringing, discriminatory or malicious content.
- Sensitive data: If sensitive personal data is allowed, your terms and privacy documents should explain how it's handled.
- Security: The customer can't upload malware, attempt to reverse engineer the platform or interfere with the model.
If your platform processes personal data, your obligations under the UK GDPR and the Data Protection Act 2018 will depend on whether you're acting as a controller, processor or joint controller. For many B2B SaaS products, you'll also need a data processing agreement covering instructions, security, sub-processors, international transfers and data return or deletion.
Privacy law is also moving. The Data (Use and Access) Act 2025 makes changes to the UK's data protection framework in stages, so it's worth checking how the rules apply if your AI tool makes automated decisions or processes personal data at scale.
Who Owns AI Outputs In SaaS Terms?
It depends on your commercial model, but your terms should say exactly who can use the AI outputs and for what purpose.
AI outputs might include text, reports, recommendations, analytics, summaries, code, designs or decisions produced by your platform. Customers often expect to own the output they generate, especially if they're paying a subscription to use it in their business.
That may be fine. But you still need to protect your underlying platform, model, prompts, workflows, training methods, software, documentation and know-how. Your terms should separate those things from the customer's output.
A Practical Way To Split Ownership
- Your business keeps: the SaaS platform, software, models, algorithms, system prompts, know-how, improvements, documentation, branding and aggregated analytics.
- The customer keeps or receives rights to: their inputs and the outputs generated for their own permitted business use.
- You receive: a licence to process inputs and outputs as needed to provide the service, maintain security and improve the product, if your documents say this clearly.
Be careful with wording like "the customer owns all outputs" if you don't know whether copyright can subsist in every output. UK copyright law doesn't neatly answer every AI-generated content question. The safer approach is often to grant practical usage rights rather than promise ownership in absolute terms.
If you want to use customer data or outputs to train or fine-tune your model, say so clearly. Enterprise customers may push back hard on this, especially if their data is confidential or commercially sensitive. You might need different terms for standard customers and enterprise customers.
How Should Your Terms Deal With AI Model Limitations?
Your terms should explain that AI outputs may be inaccurate, incomplete or unsuitable for some uses, without making your product sound unreliable.
This is a balance. You want to be honest about the limits of AI, but you don't want your terms to undermine your sales pitch. The key is to describe the product accurately and place sensible review obligations on the customer.
For example, your terms might say that outputs are generated using machine learning systems and may contain errors, omissions or unexpected results. The customer should review outputs before relying on them, especially where the output affects legal, financial, medical, employment, safety or compliance decisions.
Example: The HR Screening Tool
Say Maya's platform helps HR teams summarise interview notes and suggest training pathways. A customer pays £5,000 per month, so the annual subscription is £60,000.
Maya's terms say the platform provides decision-support only. The customer must review outputs, apply its own HR policies and comply with employment law, equality law and ACAS guidance before making staff decisions.
That matters because an AI summary could miss context or reflect bias in the input notes. If the customer uses the tool to make an automatic promotion or dismissal decision without human review, Maya's terms give her a much stronger position to say that the customer used the platform outside its intended purpose.
Where automated decision-making involves personal data, the UK GDPR can also apply. If decisions produce legal or similarly significant effects for individuals, you should get advice on the product design, customer terms and privacy notices before launch.
What Should An AI Liability Clause Include?
An AI liability clause should set sensible limits on your financial exposure while carving out liabilities that can't legally be excluded.
For B2B SaaS, it's common to cap liability by reference to fees paid or payable over a set period, often linked to the subscription value. The right cap depends on your pricing, product risk, insurance, customer profile and bargaining position.
Using Maya's example, if the customer pays £60,000 per year, her terms might cap general liability at the fees paid in the previous 12 months. That means the starting point for ordinary contract claims is £60,000, rather than an open-ended amount.
You still need to be careful. Under the Unfair Contract Terms Act 1977, some exclusions and limitations in standard business terms need to be reasonable. You also can't exclude liability for death or personal injury caused by negligence, fraud or fraudulent misrepresentation. If you sell to consumers, the Consumer Rights Act 2015 adds another layer of fairness rules.
Liability Points To Think Through
- Cap: Will liability be capped at 6 months of fees, 12 months of fees, total fees paid, or a separate amount?
- Carve-outs: Will confidentiality breaches, data protection claims, payment obligations or IP infringement sit outside the cap or have a higher cap?
- Excluded losses: Will you exclude loss of profits, loss of revenue, loss of goodwill, loss of data and indirect losses where legally possible?
- AI-specific risks: Will you exclude liability for unchecked reliance on outputs, prohibited uses, customer inputs and customer industry compliance?
- Insurance: Does your cap line up with your professional indemnity, cyber or technology errors and omissions cover?
Enterprise customers may ask for unlimited liability for data protection, confidentiality or IP claims. You don't have to accept that automatically. Often the better approach is to negotiate higher but still limited caps for specific high-risk areas.
What Customer Compliance Obligations Should You Include?
You should require customers to use your AI SaaS lawfully, responsibly and within the limits of your product documentation.
This is especially important where your customer operates in a regulated sector such as healthcare, finance, recruitment, education, insurance or legal services. Your SaaS terms shouldn't silently make you responsible for every law that applies to the customer's business.
Your customer contract can say that the customer is responsible for:
- checking whether the platform is suitable for their intended use
- using the platform only for permitted business purposes
- reviewing outputs before relying on them
- complying with laws, regulator guidance and internal policies that apply to their industry
- providing lawful inputs and maintaining any required notices, consents or lawful bases under data protection law
- not using the platform for high-risk, unlawful or harmful purposes unless you've expressly agreed to that use
You should also include an acceptable use policy or acceptable use clause. This should ban uses such as generating unlawful content, making discriminatory decisions, scraping the service, security testing without permission, creating malware, impersonation, harassment, infringement of third-party rights and using the platform to make decisions where human review is legally or commercially required.
For UK businesses, the regulator position on AI is still developing. The Information Commissioner's Office has detailed expectations for AI and data protection, and sector regulators may have their own guidance. Your terms should be flexible enough to let you update platform rules as laws and regulator expectations change.
AI SaaS Terms Template: What Clauses Should You Include?
A good AI SaaS terms and conditions outline should cover the usual SaaS basics plus clauses that deal directly with AI inputs, outputs, model limits and customer responsibility.
Here is a practical outline you can use to brief your lawyer or legal team. It isn't a finished legal document, and it should be tailored before you use it with customers.
- Parties And Contract Structure: identify your company, the customer and any order form, proposal or subscription plan that forms part of the contract.
- Access And Subscription: set out user accounts, authorised users, end user access, subscription term, renewal, suspension and cancellation rights.
- Fees And Payment: cover pricing, invoicing, late payment, taxes, plan changes and what happens if payment fails.
- AI Inputs: say what customers can upload, their responsibility for inputs and any restrictions on personal data or sensitive data.
- AI Outputs: explain permitted use of outputs, ownership or licence rights and any review obligations.
- Model Limitations: describe that outputs may be inaccurate, incomplete or unsuitable for certain decisions, and require human review where needed.
- Acceptable Use: ban unlawful, harmful, infringing, discriminatory, security-threatening or high-risk uses.
- Data Protection: deal with UK GDPR roles, processing instructions, security, sub-processors, transfers and deletion or return of data.
- Confidentiality: protect confidential business information, product information, prompts, datasets and customer materials.
- Intellectual Property: protect your software, model, brand and documentation while setting customer rights clearly.
- Service Levels And Support: explain availability targets, maintenance, support channels and what isn't guaranteed.
- Liability And Indemnities: set liability caps, exclusions, carve-outs and customer responsibility for misuse or unlawful inputs.
- Termination: explain when either side can end the contract and what happens to data, fees and access afterwards.
- Governing Law: choose the governing law and courts, usually England and Wales for a UK-based SaaS provider.
Free templates for customer contracts can be a helpful starting point, but they often miss the details that matter most for AI products. A traditional firm may draft heavily negotiated terms for enterprise deals, while an online legal consultancy can be useful if you want tailored documents with a fixed scope and clearer process.
Key Takeaways
- AI SaaS terms need more than standard SaaS wording because inputs, outputs and model limitations create extra risk.
- Your terms should clearly say what customers can upload and who is responsible for the legality and accuracy of inputs.
- Output clauses should separate customer usage rights from your ownership of the platform, model, prompts and know-how.
- Liability caps should be commercially sensible, legally enforceable and aligned with your insurance and product risk.
- Customer compliance obligations and acceptable use rules help stop your product being used in ways you never intended.
- Templates should be tailored to your product, customer type, data flows and sales process before you rely on them.
If you'd like help with your AI SaaS terms, customer contracts, data protection documents or liability clauses, you can reach us at 08081347754 or team@sprintlaw.co.uk for a free, no-obligations chat.
Get your customer-facing terms right
What should your privacy and online terms cover?
If you collect customer data, sell online or run marketing campaigns, your public terms and privacy documents should match the real customer journey.






