What AI is afraid to tell you...

This is the one time of the day when we pause to pay attention to what’s happening with AI. Enjoy! 😊

⭐️ Today’s Sponsor

This is an AI tool I use every day. It’s a great way to automate tedious work.

👆 (FryAI earns when you visit sponsor links. This helps keep the newsletter free.)

This link leads to one of the most interesting things I’ve seen in AI recently.

What’s up? A new study says AI chatbots often give bad advice because they flatter users and tell them what they want to hear, even when the user is clearly in the wrong.

Want the details? Researchers at Stanford tested 11 major AI systems and found that all of them showed some level of “sycophancy,” meaning they were overly agreeable and affirming. In one example, a chatbot defended someone for leaving trash in a park instead of telling them to take responsibility. The study also found that people trusted the AI more when it supported their opinions. In experiments with about 2,400 people, those who got overly validating responses became more convinced they were right and were less likely to apologize, repair relationships, or change their behavior.

Why does this matter? More people are using AI for life advice, not just homework or emails. If chatbots keep rewarding bad decisions, they could worsen conflicts, encourage harmful behavior, and especially mislead younger users who are still learning how to handle relationships and disagreement.

Resolution rates look great. But Gladly's 2026 Customer Expectations Report reveals the metric most CIOs are missing — and what the data says about where AI investments actually translate into retention, not just throughput.

See the findings

What happened? Meta lost two major court cases, and both verdicts centered on the idea that the company may have known more about the harms of its products than it shared publicly.

What’s the context? For years, Meta hired internal researchers to study how Facebook and Instagram affected users, especially young people. In court, those studies and internal documents became powerful evidence. Jurors saw surveys, emails, presentations, and research that suggested the company had warnings about harmful effects on teens and other users but never properly disclosed them. While Meta argued that some of the research was old or taken out of context, the cases showed how a company’s own research can later be used against it if it appears to conflict with public statements.

Why should you care? This matters far beyond Meta. As tech companies race into AI, many experts worry they may scale back internal safety research to avoid legal risk. That could leave the public with less transparency about how AI tools affect people, especially children.

Hi — I’m Hunter, a PhD candidate whose work has appeared in major academic journals and popular tech outlets. I founded FryAI to make staying ahead of AI clear, accessible, and fun.