ChatGPT use tied to rise in reports of ‘ritual abuse’, say UK experts

This story captures a strange, unsettling edge of generative AI: people are turning to chatbots not just for homework and emails, but to process deep trauma and decide whether to disclose it. Survivors of extreme abuse are apparently using ChatGPT as a low‑friction, anonymous sounding board, and sometimes that nudges them toward formal reporting and support. That’s a radically different role for these systems than anything in their original product roadmaps.

From an AGI perspective, it underscores that even “general” assistants quickly become quasi‑therapeutic in practice. As models get more capable and more emotionally convincing, the line between chatbot and counsellor blurs, whether regulators like it or not. That raises hard design questions: how much triage logic, risk assessment and referral infrastructure should be baked into frontier models by default? And how do you prevent them from giving dangerously bad advice in high‑stakes situations while still allowing them to be useful confidants?

Strategically, this also shows why safety debates can’t be siloed into long‑term existential risk. Very near‑term harms—from mis‑handled disclosures to suggestive content in vulnerable chats—are going to shape public trust, regulatory pressure, and ultimately the operating space for AGI labs.