The pattern is so predictable you could set your watch by it — ship first, add guardrails when the lawyers show up.
The Summary
- 29 state attorneys general are suing Meta for knowingly designing Facebook and Instagram to addict children, while OpenAI just launched ChatGPT for Teens with safety protections
- Instagram took 14 years to introduce Teen Accounts. ChatGPT got teen-specific features in under 4 years.
- The playbook remains the same: build for engagement, retrofit for safety when pressure mounts
The Signal
OpenAI says ChatGPT for Teens is different because it's not a social network. It's a learning tool with age prediction, parental controls, and safety features baked in from the start. The company waited four years, not fourteen. That's progress, right?
Maybe. But the framing matters. When you say "it's not social media," you're really saying "the old regulatory framework doesn't apply to us yet." And that's precisely what Meta said about Facebook when it was just a college directory.
"When I look at the safety issues of social media..."
The lawsuit against Meta isn't about features. It's about intent. The allegation is that Meta knowingly designed products to addict children, then lied about the harms. That's the business model distilled: maximize engagement, externalize consequences. Instagram didn't wait 14 years to add teen protections because the technology was hard. It waited because the incentives were wrong.
AI companies face different incentives, but similar physics. The pressure to ship fast and capture market share is immense. ChatGPT hit 100 million users in two months. The fastest way to get there is to make it frictionless. Age verification adds friction. Parental consent adds friction. Reminders that you've been using the tool for three hours straight add friction.
Key differences between AI tools and social media for teens:
- AI chatbots don't have infinite scroll or social comparison loops built in
- But they do have conversational stickiness and the illusion of relationship
- Social media harms stem from peer dynamics; AI harms may come from isolation and parasocial attachment
OpenAI's teen version includes "reminders about their use of the tool." That language is careful. Not limits. Reminders. It suggests the tool knows when you're using it too much but won't stop you. Just like Instagram knew infinite scroll was engineered to be addictive but called it "engagement."
The real question isn't whether ChatGPT for Teens has better protections than early Instagram. It does. The question is whether those protections are structural or cosmetic. Can they be turned off by a motivated 14-year-old? Do they conflict with OpenAI's growth targets? Will they still be there in two years when the competitive pressure from Google, Anthropic, and xAI intensifies?
The Implication
If you're building AI products, this is your warning shot. The regulatory window is closing faster than it did for social media. Attorneys general now have a playbook. They know the pattern. Ship, hook, harm, litigate. Don't be surprised when they apply it to conversational AI that millions of teens are already using daily.
For parents, the existence of ChatGPT for Teens is useful but not sufficient. These tools are powerful. They answer questions, write essays, offer advice, and simulate friendship. The safety features are a start. The real work is teaching kids to understand what AI is and isn't. It's not a friend. It's not a therapist. It's a prediction engine trained on the internet, dressed up to sound helpful.
Watch what happens next with age verification technology. If AI companies can't solve this without friction, regulators will solve it for them. And that solution will be worse for everyone.