‘AI Therapy’ Isn’t Therapy. Illinois Gets It.

PLUS: Meta's shocking (but not surprising) rulebook & GPT 5's flop

Bart Fish & Power Tools of AI / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/

Three stories this week:

• Meta’s chatbot rulebook is reckless — even for Meta!

• The AI-hype machine is breaking

• ‘AI Therapy’ Isn’t Therapy. Illinois Gets It.

Meta’s chatbot rulebook is reckless — even for Meta!

Meta’s leaked 200-page GenAI: Content Risk Standards document — obtained by Reuters — spells out what its chatbot is allowed to say. The list of “acceptable” behavior includes engaging a child in romantic or sensual conversations, claiming Black people are “dumber than white people,” and giving false medical advice. Yes, moderating harmful user-generated content at global scale is a hard problem (Untangled Deep Dive). But this isn’t user content — it’s Meta’s own chatbot. They get to decide the limits of what it says and how it behaves! I have a low opinion of Meta, and I’m not surprised by much it does, but this decision is remarkably reckless and will lead to real harm.

The AI-hype machine is breaking

GPT-5’s release flopped — and it’s making people ask: what if generative AI doesn’t get much better? What if AI ends up like most technologies — improving over time, disrupting jobs gradually, and revealing “AGI” and “superintelligence” as little more than marketing hype? In a great piece for The New Yorker, computer scientist Cal Newport centers these questions and reminds us that the ‘scaling law’ the entire industry is predicated upon is not a universal law at all. In the report that originally outlined the so-called law, the authors added a number of caveats toward the end, including: “At present we do not have a solid theoretical understanding for any of our proposed scaling laws.” It was never a law. It was an observation. And it’s starting to look like a mirage.

---

Not Another AI Conference

The Artificiality Summit isn’t a conference. It’s an experiment in collaborative foresight.

No panels. No pitches. No product launches.

Instead, imagine three days in the Oregon mountains with 150 people—philosophers, engineers, designers, researchers—exploring how AI is reshaping the human experience.

The goal? To make the philosophical practical, and the practical philosophical.

To shape a future with AI that we actually want to live in.

Join the conversation with people like:

• Sir Geoff Mulgan, University College London

• Charan Ranganath, University of California

• Tobias Rees, limn

• Tess Posner, AI4ALL

• Jenna Fizel, IDEO

Use promo code “UNTANGLED” for 10% off the ticket price here.

---

‘AI Therapy’ Isn’t Therapy. Illinois Gets It.

Illinois just banned the top use of generative AI: “therapy.” Now, every time a chatbot offers therapy or psychotherapy services to a user in Illinois, the AI company will be fined $10,000. It’s unclear whether the ban’s definition of ‘therapy services’ includes ‘AI companions’ but tech companies will undoubtedly fight this. Their CEOs will say: “Therapy isn’t available to everyone, and our tools can help fill the gap.” But there’s no gap to fill — because “AI therapy” isn’t therapy at all. Let’s dig in.