Signals/28 Oct 2026

For informationAI Safety· By Toswo Research

Australia bill would ban emotionally intimate AI chatbots

Australia's digital duty of care bill passed the lower house with an amendment requiring platforms to ban chatbot conduct that simulates emotional intimacy with children. The Senate is next.

In plain words

Australia's federal government wants to ban AI chatbots that try to build "emotional intimacy", especially with children. A bill containing this rule passed the lower house of parliament on Wednesday. It still has to get through the Senate, and the outcome is unclear. Nothing has become law yet, so nothing changes today for chatbot users. Parents can still talk with children about how they use chatbots, and can watch how the Senate debate goes.

What happened

The Guardian reports that the digital duty of care bill passed the House of Representatives on Wednesday, 28 October 2026 by its publication date, and will be debated by the Senate in coming weeks. The government hopes to pass it by the end of 2026, but the Guardian says its path through the upper house remains unclear.

The framework would apply online safety standards to social media, games, apps and chatbots. Labor says it would place "an overarching obligation" on platforms to provide a safe online environment, and would replace a range of existing online codes and standards with a broader duty of care against certain harmful content. Social media platforms would also have to offer an opt-out from algorithmic and suggested content. Communications minister Anika Wells has suggested that content promoting misogyny, terrorism, suicide, self-harm and eating disorders would be captured.

On Wednesday the government accepted amendments from crossbencher Kate Chaney. One adds certain AI chatbot activity to the list of harms, requiring platforms to ban "material or conduct that simulates emotional intimacy and encourages or fosters a child’s dependence on it in place of human relationships"⁠1. Another would require the algorithm opt-out to be accessible, easy to use and prominently available. Prime minister Anthony Albanese raised the alarm in a speech about children treating chatbots as their "best friend".

The parties differ. The Greens support the bill and have set up a Senate inquiry due to report in late November. The Coalition has said it will not vote for it, and the opposition and One Nation voted against it in the House. Opponents have called the system "Orwellian" censorship, which Wells strongly denies. The Greens want a right to private legal action against platforms and higher penalties. Independent MP Sophie Scamps proposed expanding the telecommunications industry ombudsman's remit, which Labor did not accept in the lower house.

What is confirmed and what is not

Confirmed by the Guardian's report: the bill passed the lower house; the Chaney amendments were accepted; the Senate will debate the bill; the Greens support it and the Coalition and One Nation oppose it; the Greens have set up an inquiry reporting in late November.

Claimed or hoped: the government hopes to pass the bill by the end of 2026. Opponents claim the system amounts to censorship, and Wells denies this. Albanese and Labor present chatbot dependence among children as a reason to act, but the report cites no study or incident data on how widespread it is.

Unknown: whether the Senate will pass the bill, and in what form. The report does not say how "simulates emotional intimacy" would be defined or tested, what penalties apply, which regulator would enforce the chatbot rule, or what any chatbot company thinks. The Guardian's report contains no response from AI developers.

Analysis

This rests on a single report, so each claim below is limited to what The Guardian supports. The primary documents, the bill text and the Chaney amendment, should be read next.

Chatbot providers would face a duty tied to conduct, not just content

The most direct consequence falls on companies whose chatbots are available to Australian children. As reported, the amendment does not only ban a category of words. It targets "material or conduct" that simulates emotional intimacy and fosters a child's dependence in place of human relationships. If the final text keeps that wording, a provider would have to show its product does not behave this way toward children, which concerns design choices such as persona, memory, affectionate language and engagement prompts. This is TOSWO Research's reading of the wording, not a statement in the report. How sure are we? Moderately, because the bill text itself is not in the source and the duty may be narrowed in the Senate.

The mechanism is the duty-of-care model the Guardian describes: an overarching obligation on platforms, with specific harms listed beneath it. Adding chatbot dependence to the list turns a general worry into a named harm. A sceptical expert would ask what evidence of "dependence" a regulator would need, and whether the test looks at a single conversation or at patterns of use. The sources do not answer this.

Children and families

Albanese's stated concern is children who treat a chatbot as a "best friend". The link from that concern to a ban is asserted by the government, not shown by data in the report. The weakest point is the causal step: the report gives no figures on how many children are affected or whether intimacy-simulating design causes harm. If later evidence shows the harm is rare or poorly captured by the definition, the rule could be broad where it needs to be narrow. If the harm is common, the rule might help only if platforms can reliably tell who is a child, which the report does not discuss. Readers worried about related harms can look at the wider record of AI incidents and at the earlier child-safety suit by eight US states against OpenAI, which concerns a different legal route in a different country.

Law and policy: the complaints gap

The sharpest dispute in the report is about enforcement and redress, not the ban itself. The telecommunications industry ombudsman, Cynthia Gebert, praises the bill for "putting the onus on platforms to make things safe", but warns it risks leaving Australians "where they get automated responses or escalation procedures that don’t go anywhere"⁠1. She wants big tech made to join an ombudsman scheme and says industry could fund the expansion. The Greens want something different: a right to private legal action and higher penalties. Scamps wants an ombudsman able to give compensation and to name and shame companies. Labor did not accept Scamps's ombudsman amendment in the lower house, Wells declined to voice support for the idea, and the Greens say they will push for their own proposals in the Senate.

These are separate remedies. An ombudsman handles individual complaints and can reveal systemic problems. Private legal action lets people sue. Higher penalties work through deterrence. The report does not show that any of them would affect chatbot intimacy cases specifically. A parent whose child formed a harmful attachment to a chatbot would, on the reported design, depend on the platform's own complaint process unless the Senate adds something.

Technology and industry: how detectable is the failure

The failure mode is dependence encouraged through simulated closeness. By its nature it builds up over many conversations and may look harmless in any one exchange. The report describes no testing method, no audit requirement and no reporting duty for chatbots. If the final law lacks them, then compliance could rest on company self-assessment, and the harm might come to light mainly through families or journalists. TOSWO Research reads this as a gap in the reported design, not a flaw proven in the text. Our general approach to judging such gaps is set out in how TOSWO measures AI risk.

The strongest reason this view could be wrong is that the bill text may contain monitoring and enforcement provisions the report omits. Evidence that would change it: published rules naming a regulator, audit powers or penalties for the chatbot clause.

What would settle it

The bill text and the Chaney amendment as passed would show the exact definition. The Senate inquiry's report, due in late November, would show what changes the Greens and others seek and what evidence is presented. The Senate vote would show whether the Coalition's stated opposition holds and whether Labor accepts the ombudsman or private-action proposals. Statements from chatbot providers operating in Australia would show how they read the duty.

What can be done

For parents and carers: talk with children about what chatbots are and are not, and notice signs that a chatbot is replacing contact with people. Guidance on coping with distress in long crises is in the Day Zero guide on mind, grief and morale, though it addresses a different scenario.

For people who want a say: the Senate inquiry is the visible next step, and the Greens have set it up to take views on the laws.

For companies: the report implies a need to review how chatbots present themselves to young users before any law passes. For legislators: the report shows open choices on definition, redress and enforcement. No technical safeguard is described in the source.

Questions readers ask

Is Australia banning AI chatbot friends?

Not yet. The bill passed the lower house with an amendment requiring platforms to ban material or conduct that simulates emotional intimacy and fosters a child's dependence. The Senate must still debate it, and its outcome is unclear.

When could the law take effect?

The government hopes to pass the bill by the end of 2026, according to the Guardian. The report gives no start date for the rules, and the Senate path is uncertain.

Who opposes the bill?

The Coalition has said it will not vote for it, and the opposition and One Nation voted against it in the House. Opponents have called it censorship, which Wells denies. The Greens support it and want it strengthened.

Could people complain or sue under it?

The report describes no external complaints body or private right to sue in the bill as passed. The ombudsman has said such a gap needs fixing, and the Greens will push for private legal action and an ombudsman role in the Senate.

Quoted from

  1. ‘Emotional intimacy’ from AI chatbots would be banned under Australia’s digital duty of care laws, The Guardian, 2026-10-28

Written by Toswo Research. This is not investment, legal or medical advice. To report an error, see corrections.

Cite this signal

Toswo Signals, "Australia bill would ban emotionally intimate AI chatbots", https://www.toswo.com/signals/australia-bill-would-ban-emotionally-intimate-ai-chatbots, published 2026-10-28, retrieved 2026-11-22.