Station 15· What you do with it
Can an AI chatbot talk to my customers, and what can I say about it?
This is the first leg back in your hands: the answer goes out under your name, so what customers are told, and what an ad claims, become yours to answer for.
General information from Australian Government guidance on AI tools, not legal advice.
Watched, and unwatched
The government's own pair of cases is about an AI assistant with a door to the public, and it comes from its guidance on AI adoption. A chatbot answering simple questions while the business is open, with a staff member able to watch it, is what the guidance calls a low-risk use of AI. The risks expand when the same chatbot works around the clock with nobody overseeing it, and takes on more complex questions.
The same chatbot appears in both cases, and what differs is the watching, so the watching is the control. The guidance asks for oversight that is meaningful: a person watching the system in proportion to how much freedom it has and how high the stakes are. And it asks for the points built in, explicit places where a person can pause the system, override it, roll back a change or shut it down when needed.
Read as a pair, the two cases answer a question a business actually has: not whether a chatbot may talk to customers, but what has to be true when it does. In the government's example, the low-risk version has a staff member within reach during the hours the business is open, and the risky version has nobody, all night, on harder questions.
Customers should know it is AI
People should know when they are talking to AI, and understand when AI decisions affect them, and a customer getting services and guidance from a chatbot should know it is not a human specialist.
The privacy basis for that line is the OAIC's. As part of an entity's APP obligations on transparency, and on lawful and fair collection, the OAIC says an entity should make sure people are told they are dealing with an AI system and not a human. The adoption guidance carries the same instruction, in almost the same words, for a customer getting services and information from a chatbot.
One source is advisory and one is a regulator, and they agree on the direction of travel: the person on the other side of the conversation is told what they are talking to, and the business is the one that tells them.
What your ads and listings may claim
The consumer law arrives with the advertising. The National AI Centre gives the example in one line: misleading statements made by AI-driven advertising or promotions could violate consumer laws.
The ACCC's snapshot names the practices it is watching. On its list of potential consumer issues are the widespread use and collection of consumers' data, the use of AI to facilitate false representations, AI-generated volumes of fake reviews, and AI that facilitates and enhances online scams. Its chair names three practices in her own words. Ghost websites misrepresent themselves as local businesses, and often use generative AI images to build a sense of credibility. Online product listings may use generative AI to make products appear more sophisticated, or of a higher quality, than they actually are. And AI may be used to generate and disseminate large volumes of fake reviews, which may be seen as more credible and persuasive by consumers, and be increasingly more difficult to detect.
For businesses in its field, one more regulator states the responsibility in a single sentence: the Therapeutic Goods Administration's guidance on advertising on social media says business owners and advertisers carry responsibility for AI-generated or AI-assisted content they manage and disseminate, including on social media.
Who answers when it goes wrong
The line that closes the loop comes from the government's own guide to choosing an AI solution: whichever solution a business chooses, the business remains responsible for how AI is used.
The adoption guidance says the same about the whole system. Overall, an organisation is ultimately accountable for how and where AI is used, and AI complexity can create gaps where no one takes clear responsibility for outcomes.
Put beside the sections above, the sentence is not abstract. The telling in the second section, the claims in the third, and the watching in the first are all the business's own work, and an assistant carries none of them on the business's behalf. That is why this leg of the round trip is drawn in plain ink on the site's home page: the answer has come back, and it is in your hands.
The whole round trip, station by station (aipersonalassistant.com.au.html)