When Your AI Chatbot Misleads a Customer, You're Liable

Transcript

Imagine waking up and finding out your plumbing business is on the hook for a massive commercial job. Oh, at like a 90% discount. Yes, exactly. And all because your brand new AI phone agent just hallucinated a promotional offer while you were asleep. Yeah, it totally turns that dream of having your missed calls answered 24-7 into, well, a massive liability. It really does. So if you are a time-poor tradie, you know, an electrician, a builder, or maybe you're running a small professional services firm like an accounting practice here in Australia, you need to listen up. For sure. Because today we're taking a deep dive into a critical Codery Labs article. We're looking at this exact hidden legal landmine that's just waiting for small businesses using these AI chatbots. Right, because, I mean, we all want that workflow efficiency, but that unpredictability with AI, it isn't just a customer service headache anymore. No, it's much worse. Yeah, if your bot promises something you can't actually deliver, you haven't just annoyed a client, you've actually broken Australian consumer law. Right, specifically Section 18, which is the one that prohibits misleading or deceptive conduct. And the Treasury actually looked at this in 2025. They confirmed our current laws are perfectly fit for purpose to handle AI. And the scary part, intent is entirely irrelevant here. Exactly. The law essentially views your AI as your legal agent. Wow. Yeah, because you programmed the parameters, you know, you plugged it into your phone line, you put it on your website. Right. You gave it the authority to speak on your behalf. So, well, you own its mistakes. So, it's basically like sending a first-year apprentice out to, quote, a major commercial job, completely unsupervised. Yes. Except, you know, they're just wildly guessing the prices. That analogy is spot on. The law treats what that AI says exactly as it would treat you or your staff saying it. Wait, really? But if the AI vendor sold me a faulty bot that just invents a fake refund policy out of nowhere, can I just blame them? You'd think so, right? Yeah, like point the finger at the software company for giving me a defective product. I mean, you can try, but no. Because the customer is dealing with your business, not the software provider. Ah, right. And if you actually read the fine print, most AI vendor contracts are specifically drafted to cap their own liability. Not to catch yours. You deployed the bot, so you take the risk. Man, and the penalties for taking that risk blindly are terrifying. No, they're massive. Yeah, the ACCC, you know, our national consumer watchdog, they've officially named AI-enabled practices in their 2026 to 2027 enforcement priorities. Which is a huge red flag. Totally. Yeah. The maximum corporate penalty for a breach is, get this, the greater of $100 million, three times the benefit obtained, or 30% of your turnover. Which would bankrupt most small businesses instantly. And, you know, we already have precedent for how tributals view this stuff. Oh, right, the airline case. Yeah, in 2024, a tribunal actually forced Air Canada to honor a fake bereavement fair policy. Their own chatbot had completely hallucinated it. Unbelievable. And the airline tried to argue, oh, the chatbot is a separate entity. It's responsible for its own actions. I can't imagine a tribunal buying that for a second. No, they rejected it outright. The ruling basically established that the bot is a part of your business systems. End of story. Wow. Okay, so if we can't just delete the bot. Because we still need that efficiency. Right. We still need that 24-7 lead capture, so we have to, like, put a leash on it. I'm guessing that means restricting it to basic FAQs. Yeah, that is the practical step the CoterieLabs article recommends. You need to draw a bright line between information and commitment. Okay, so what does that look like in practice? Well, you let the AI handle the factual low-stakes questions. You know, your opening hours, your service areas, what to generally expect on a callout. Right, the simple stuff. Exactly. But the moment the conversation crosses into a binding representation. Like quoting a firm price or confirming a booking. Yeah, or processing a refund. You route that either to a human or to a strictly verified script that actually matches your real policies. Ah, I see. So you get to keep the 24-7 lead capture without signing up for the 24-7 liability. That's the goal. It's like installing a faulty safety switch, right? Huh. The thing designed to protect your time is actually what burns the house down if you don't wire it correctly. Great way to put it. So we keep the bot as the friendly receptionist, but we never let it be the deal closer. Ultimately, yeah. Because claiming the AI said it, well, that's just not going to save you in a tribunal. Not at all. But, you know, this raises an important question for you to take away and think about. Oh, what's that? Well, if your AI is out there confidently hallucinating wild policies, incorrect prices, or impossible promises to your customers, what does that actually reveal about the underlying gaps in the training data and internal policies you gave it in the first place? Oh, wow. Yeah, that is definitely something to think about before you let your newest digital apprentice answer the phone. For sure. Well, thanks for joining us on this deep dive. Get you next time. That was a Field Note brought to you by CoterieLabs. We help Australian businesses work out where AI actually pays and where it doesn't. The full write-up with every source and link is in the show notes. And if you're wondering where AI fits in your own business, there's a free two-minute scan at CoterieLabs.com.au. New Field Note. Every weekday. Thank you.