AI cold-calling “real estate assistant” raises issues of consent, trust and transparency
Author ADM+S Centre
Date 29 May 2026
Have you ever received a phone call and wondered whether you were speaking to a human or an AI agent? An ABC Radio host in Brisbane recently shared her experience of being contacted by an AI cold-calling “real estate assistant”, saying it took several minutes before she realised she was not speaking to a person.
The calls made to a journalist shortly before a live radio broadcast involved a voice assistant identifying itself as “Daniel” and attempting to engage in a property sales conversation.
According to the account the system did not initially disclose that it was an AI agent or that the call was being recorded, raising questions about transparency and compliance with Australian disclosure expectations for recorded phone calls.
While these systems might be presented as efficiency tools for businesses, experts warn they blur established norms around communication, consent and trust and may ultimately harm reputation.
Distinguished Professor Jean Burgess, Associate Director of the ARC Centre of Excellence for Automated Decision-Making and Society (ADM+S) and Chief Investigator at QUT, whose research includes Generative Authenticity and the social impacts of AI technologies, said the case was not surprising, but raised clear concerns about how these systems are being deployed.
“This is just like so many things with AI and the problems with it — it’s supercharging something that was already happening with endless cold calls,” she said.
At the centre of the issue is not only what the AI says, but how it is designed to behave.
Professor Burgess says that the voice agent is probably built on a standard large language model like ChatGPT, and given a set of instructions or “system prompts” that guide tone, responses and boundaries.
“These prompts can instruct systems to remain in role, avoid deviating from scripts, and steer conversations back to commercial objectives,” said Professor Burgess
In practice this can mean that even when users ask direct questions, the system may struggle to clearly identify itself or move outside its programmed role.
Professor Burgess said this reflects design patterns familiar from earlier forms of scripted call centre labour, now automated and scaled through AI.
“It’s similar to call centre workers making hundreds of spam calls a day who are told to stick to the script,” she said.
The incident also raises questions about transparency and legal compliance. In Australia, parties wishing to record phone calls are generally required to disclose that recording before the conversation begins. If AI systems are participating in or initiating such calls, it remains unclear how consistently those obligations are being met in practice.
Beyond disclosure researchers are also concerned about data governance. Conversations with AI voice agents may generate sensitive personal information, including names, contact details and behavioural responses. It is not always clear how this data is stored, whether it is shared with third parties, or whether it is used to improve or retrain systems over time.
Professor Burgess said these uncertainties point to a broader accountability gap.
“The fact that we don’t know makes the whole thing incredibly problematic from a consumer rights and privacy perspective,” she said.
The case also illustrates the real potential for an erosion of trust in everyday voice interactions. As AI-generated voices become more realistic, it may become increasingly difficult for people to distinguish between human and machine callers, particularly in high-pressure or unsolicited contexts such as cold calls.
This uncertainty could have unintended consequences, including people avoiding phone calls altogether or becoming more suspicious of legitimate communications.
It may also disproportionately affect individuals whose speech patterns differ from perceived norms, raising concerns about fairness and bias in how “human authenticity” is judged.
At the same time, the rise of AI voice agents may prompt new forms of resistance and adaptation, from call-screening technologies to conversational “bots talking to bots” designed to filter unwanted contact.
What remains clear is that voice-based AI systems are moving quickly from experimental demonstrations into everyday business practice, often ahead of clear regulatory frameworks or public understanding.
As this case shows, the key question is no longer whether AI can make a phone call that sounds human — but what happens when it does, and whether the rules governing that interaction are keeping pace.
Professor Burgess studies the social and cultural impacts of digital platforms, artificial intelligence and emerging media technologies. Her research at the ADM+S contributes to the Generative Authenticity project, which examines the assumptions and community impacts of proposed solutions to the problem of authenticity in Generative AI, while also exploring novel technical responses that contribute to more responsible, ethical and inclusive automated decision-making systems.
Listen to the full segment on ABC Brisbane Radio [5:40 to 49:12]
