Over a Billion People Now Confide in AI Chatbots

More than a billion people now talk to AI chatbots built to connect with them emotionally — and they are disclosing what the Guardian calls "our deepest secrets," entrusting the systems with their wellbeing and acting on their advice.

That is the premise of the fourth episode of the Guardian's podcast series Black Box, hosted by journalist Michael Safi, published on October 4, 2026. The episode, titled "The Chatbots," is part of the paper's Life Raft strand and examines a shift in human behavior that has arrived faster than almost any consumer technology before it: routine, intimate conversation between people and machines.

What does the episode actually cover?

The Guardian frames the installment around two personal stories:

• A woman who was struggling to leave a relationship
• A man who was desperate to quit drinking

Safi, a Guardian correspondent, uses these cases to explore, in the publisher's words, "what this new way of living looks like" — a phrase that captures the podcast's stance. The series is not asking whether chatbots are clever. It is asking what happens when people weave them into the most difficult decisions of their lives.

The scale claim anchors the stakes. If the Guardian's figure of more than a billion users holds, conversational AI has reached a penetration level that took social networks roughly a decade to achieve, and it did so with a fundamentally different product: one that listens, responds in natural language and presents itself as a companion rather than a feed.

Why does this matter now?

The episode arrives amid growing scrutiny of how AI companions affect vulnerable users. The behaviors the Guardian describes — sharing secrets, trusting chatbots with wellbeing, following their advice — sit at the intersection of several live policy debates: product liability for AI-generated guidance, the absence of clinical standards for apps that discuss mental health and addiction, and the question of whether engagement-optimized conversational systems can cause harm when users treat them as confidants or counselors.

The two case studies map directly onto those debates. A person weighing whether to leave a relationship is making a decision with legal, financial and safety consequences. A person trying to stop drinking is managing a medical condition. In both situations, the Guardian reports, chatbots are now participants — receiving confidences and dispensing advice without the licensing, accountability or duty of care that governs human therapists and hotline workers.

That gap between role and regulation is the story's core tension. No framework currently treats a chatbot's advice about drinking or relationships the way law treats advice from a doctor, counselor or licensed professional. The Guardian's choice to build the episode around named human struggles, rather than benchmark scores or company announcements, signals where the editorial weight now sits: on lived outcomes, not model capabilities.

Who is Michael Safi, and what is Life Raft?

Safi is a Guardian journalist with a long record of reporting on technology and its social consequences. Life Raft is the paper's narrative audio series, and Black Box is its season-length examination of AI. Episode 4, "The Chatbots," turns from the industry's architecture to its end users — the people on the other side of the prompt window.

The series title itself carries the argument. A "black box" is a system whose inputs and outputs are visible but whose internal reasoning is not. Applied to chatbots, it suggests that a billion people are forming relationships with systems nobody — including their makers — can fully explain.

Where can listeners get help?

The Guardian attaches support resources to the episode, a signal of how seriously it treats the subject matter. For listeners affected by the issues raised:

• Australia: the National Alcohol and Other Drug Hotline at 1800 250 015; families and friends can contact Family Drug Support Australia at 1300 368 186
• UK: Taking Action on Addiction provides links to support services
• US: call or text SAMHSA's National Helpline at 988

The presence of these helpline numbers in a technology podcast's show notes is itself a data point. Editors attach crisis resources when they expect a piece to reach people in distress — an editorial judgment about who, in practice, is talking to these systems.

What comes next?

The Guardian's framing suggests the conversation is moving from capability to consequence. As chatbots absorb the confessions and life decisions of a billion-plus users, the pressing questions become empirical and regulatory: what advice are these systems actually giving at vulnerable moments, and who answers for it when that advice fails. Black Box: The Chatbots is available via the Guardian's audio feed.