AI Security and ComplianceAustralian small business

Who Is Liable When AI Gives a Customer Wrong Information?

Who is liable for AI mistakes when a chatbot gets it wrong? See how the Australian Consumer Law applies and 5 ways to cut your risk.

Last reviewed: 12 November 2026

AI security and business data protection

Quick answer: In Australia, the business that deploys the AI is usually liable, not the AI or its vendor. If a chatbot misstates a customer's rights, prices or refunds, that can be misleading conduct under the Australian Consumer Law, and consumer guarantees still apply. Vendor contracts usually cap the vendor's own liability, so most of the risk stays with you.

Your chatbot tells a customer they can get a refund after 60 days. Your actual policy says 30. Who pays? It's a fair question to ask before you put AI in front of customers, and the answer to who is liable for AI mistakes is clearer than most business owners expect. Whether you're using an off-the-shelf bot or custom AI solutions, the law treats the AI as part of your business, not as a separate party you can blame.

Why the Business Usually Carries the Risk

An AI tool isn't a legal person. It can't be sued, fined or held responsible. When it speaks to a customer on your website, in your app or on your phone line, it's speaking for you, exactly as a staff member would.

Australian regulators take that view. The Australian Consumer Law (ACL) bans misleading or deceptive conduct in trade or commerce, and it's technology-neutral. A false statement about a price, a warranty or a refund is treated the same whether a person typed it or a language model generated it. The ACCC has made it clear that blaming the algorithm isn't a defence.

The consumer guarantees sit underneath all of this. If a chatbot tells a customer their faulty product isn't covered because the manufacturer's warranty has expired, it's ignoring rights the ACL gives them regardless of any warranty. That's a classic hallucination with a real legal cost. And for companies, the penalties for misleading conduct can reach the greater of $50 million, three times the benefit gained, or 30% of adjusted turnover.

Real Cases Where a Chatbot Got It Wrong

The best-known case comes from Canada. In Moffatt v Air Canada (2024), the airline's website chatbot told a grieving customer he could claim a bereavement discount after booking. That wasn't the airline's policy. Air Canada argued the chatbot was effectively a separate entity responsible for its own actions.

The British Columbia Civil Resolution Tribunal rejected that argument outright. It held the airline was responsible for everything on its website, whether the information came from a static page or a chatbot, and ordered it to pay the fare difference plus interest and fees. It's a small-claims decision and doesn't bind Australian courts, but the reasoning lines up closely with how the ACL already works here.

Now picture the Australian version. An Adelaide electronics retailer uses a chatbot to handle warranty questions. A customer with a 14-month-old laptop is told it's out of warranty and there's nothing to be done. Under the ACL consumer guarantees, a laptop is generally expected to last longer than that, so the customer may still be entitled to a repair, replacement or refund. The retailer has now made a misleading statement about consumer rights, possibly to hundreds of customers before anyone notices.

What AI Vendor Contracts Usually Say

If you're hoping the AI provider will cover you, read the terms closely. Most standard agreements push the risk back to the business using the tool. Common clauses include:

  • Accuracy disclaimers stating outputs may be wrong and must be checked before use.
  • Liability caps limited to the fees paid in the previous 12 months, or less.
  • Exclusions for consequential loss, which rules out most customer claims.
  • Responsibility for use placed on you, including how outputs are shown to end users.
  • Indemnities that run one way, often protecting the vendor rather than you.

That creates two separate conversations. One is between you and your customer or the regulator, where you're usually on the hook. The other is between you and your vendor, where your recovery depends entirely on what you negotiated.

5 Ways to Reduce Your Liability

  • Limit what the bot can answer. Keep it away from refunds, warranties, legal rights, safety and regulated advice. Route those topics to a person or to fixed, approved wording.
  • Ground answers in your own content. Connect the AI to current policy documents rather than letting it answer from general knowledge, and update the source whenever a policy changes.
  • Test before launch and keep testing. Run the questions customers actually ask, especially edge cases, and review a sample of live conversations every week.
  • Make escalation easy. Give customers a clear path to a human, and tell them they're talking to an AI.
  • Negotiate the contract. Push for higher liability caps, an accuracy commitment for your use case, and a mutual indemnity, especially for custom builds.

Notice what's missing: a disclaimer on its own. A line saying "our chatbot may make mistakes" doesn't override the ACL, and it won't fix a misleading statement after the customer relied on it.

When to Get Legal Advice

Speak to a lawyer before launch if your AI will discuss pricing, refunds, credit, insurance, health or anything that affects a customer's legal rights. The same applies if you operate in a regulated sector, or if your vendor won't move on liability terms.

Get advice quickly if a chatbot has already given wrong information to a group of customers. Acting early, by correcting the information and honouring reasonable expectations, often costs far less than a regulator complaint. In most cases, it's simply the right thing to do.

Plan your next steps with Zynex Technologies

Discuss your AI security and compliance needs

Talk to the Zynex Technologies team about your business and its AI systems.

Book a call