It might seem easier than ever to start and run a business using artificial intelligence (AI).<\/p>\n
At the extreme end, some AI tools offer to run key business functions \u201cwhile you sleep<\/a>\u201d, or even build entire apps<\/a> from a single prompt. <\/p>\n But this technology carries new risks and is reshaping some old ones. And we aren\u2019t just talking about how \u201cvibe coding<\/a>\u201d \u2013 creating something like an app from scratch, without needing any coding skills \u2013 might result in glitches for customers.<\/p>\n No matter how big your business, the key customer protections<\/a> in the Australian Consumer Law still apply. Businesses are also bound by their own contractual obligations to deliver what they promised on time, in the way they promised. <\/p>\n If you\u2019re running or looking to start a business using AI tools \u2013 from writing your marketing text, to AI chatbots interacting with your customers \u2013 here are some of the key legal risks you need to manage. <\/p>\n When a chatbot gets it wrong<\/p>\n Australian law has well-established protections against misleading conduct. While it is a new technology, there is no \u201cfree pass\u201d for disruption caused by a rogue AI<\/a> instead of a human being.<\/p>\n Canada, where similar protections apply, offers a now infamous example<\/a>. In this case, an Air Canada customer assistance chatbot provided incorrect information about the process for requesting \u201cbereavement\u201d fares, leading to the customer proceeding to pay for a full fare.<\/p>\n In 2024, a tribunal found this information was misleading and ordered the airline to pay compensation. <\/p>\n There are several other ways an AI chatbot used by a business could accidentally mislead consumers. Importantly, liability does not depend on this business intending to mislead, or even being careless.<\/p>\n For example, an AI pricing tool that means some consumers do not receive a discount represented as available to them would be misleading. <\/p>\n And a chatbot that manipulates consumers\u2019 emotional responses \u2013 whether by flattery, emotional targeting or deceptive design \u2013 could fall foul of the existing ban on unconscionable conduct<\/a> and the new ban on unfair trading<\/a> that comes into effect next year.<\/p>\n It\u2019s not just chatbots; marketing is another place businesses are exposed to AI risks.<\/p>\n Many businesses are now using AI to create social media content and are likely responsible for images that are inaccurate, either in the way they present the product being sold or suggesting the social media influencer promoting it is human<\/a>. <\/p>\n Managing paperwork, quotes and orders<\/p>\n AI can also be rolled out to assist a new business with managing paperwork and orders. Errors \u2013 such as misquoting the price<\/a> of a product \u2013 may be costly.<\/p>\n And while they may be able to improve efficiency, these activities are still bound by the law, including meeting Corporations Law obligations<\/a> relating to reporting and governance.<\/p>\n The performance of the AI needs to be overseen by a human with the right expertise.<\/p>\n Managing artificial employees<\/p>\n Using AI \u201cagents<\/a>\u201d poses other risks too. These are AI systems that act with a high level of autonomy to complete discrete tasks, as opposed merely to generating content.<\/p>\n Emerging uses include retail sales<\/a>, managing merchandise distribution<\/a>, and engaging in high-frequency crypto trading<\/a>. <\/p>\n Agentic AI is \u201cmulti-modal\u201d, meaning it interacts with external tools, databases and even other AI agents to complete tasks. <\/p>\n Using AI agents or agentic AI may have real-world stakes for a business\u2019s money, inventory or more.<\/p>\n Any agentic AI system that messes up an order \u2013 to consumers or another business \u2013 still leaves responsibility with the party contractually obliged to deliver the product or service. <\/p>\n A business that rolls out an underperforming AI may itself suffer losses. These could include having to pay compensation to customers, through to facing fines from regulators. <\/p>\n Who is responsible?<\/p>\n So what if things do go wrong with the AI tools you\u2019ve used? Could your business seek compensation from the developer or tech firm that supplied that AI product? And what about all the other companies involved in developing that product? Are they legally responsible too?<\/p>\n The answer is: it depends. <\/p>\n