Your business is probably using AI in more places than you think. A chatbot answering customer queries. A model scoring applications or flagging risk. An automation tool making decisions that used to sit with a person.
Most of that happened quickly, and quietly. Few businesses stopped to ask the question that actually matters: when the AI gets it wrong, who is legally and commercially responsible?
The answer is rarely straightforward, and it is almost never "the AI provider." Here is what actually determines where the liability sits, and what to check before it becomes a problem.
Why This Is Different From Ordinary Software Risk
With traditional software, the logic is fixed. It does what it was built to do, and if it fails, you can usually trace the fault back to a specific defect or decision.
AI tools, particularly those built on large language models or machine learning, behave differently. Outputs can vary, degrade, or be wrong in ways that are hard to predict and harder to explain after the fact. That unpredictability does not remove your liability. It just makes it harder to see coming.
Where The Risk Actually Sits
In most cases, the business using the AI tool carries far more exposure than it expects. Here is why:
- Your customer terms, not the AI provider's terms, govern your relationship with your customer. If the AI gets something wrong in front of a client, your customer looks to you first.
- Most AI providers exclude liability for output accuracy in their terms and conditions. Read the fine print and you will usually find phrases like "provided as is" or "no warranty as to accuracy."
- Liability caps on AI-as-a-service contracts are often set low, sometimes at a single month's subscription fee, regardless of the scale of the mistake.
- Ownership of the AI's output is frequently unclear, which becomes a real problem if that output ends up in a customer deliverable, a contract, or a regulatory filing.
None of this means AI tools are too risky to use. It means the contract terms around them need to reflect how the tool is actually being used in your business, not how they were used when the terms were written.
Five Things To Check Before You Rely On An AI Tool Commercially
- Who signed off on using this AI tool for this specific purpose, and is that decision documented anywhere.
- What the supplier's contract actually says about liability for incorrect, biased, or harmful output.
- Whether your own customer-facing terms mention AI use, and whether they should.
- Who owns the output the tool produces, particularly where that output is shared externally.
- What your process is if the tool gets something wrong in front of a client, a regulator, or the board.
If you cannot answer most of these confidently, the AI tool is not necessarily the problem. The contract around it usually is.
Regulation Is Catching Up, But Slowly
The EU AI Act introduces obligations depending on how an AI system is classified and used, and UK businesses trading into the EU, or using AI systems built by EU-based providers, need to understand where they sit within that framework. Regulation in this area is still developing, and it will not resolve the contractual gaps that already exist today.
Waiting for regulatory clarity is not a strategy. The commercial exposure exists now, regardless of what the rules eventually say.
The Bottom Line
AI tools move fast. Legal frameworks, by nature, do not. The businesses that get caught out are rarely the ones taking the biggest risks with AI. They are the ones who assumed the contract already covered it.
If your AI tools have moved faster than your legal framework, book a call with the Ethiqs team