According to Gizmodo, in a letter dated December 9 and made public on December 10, dozens of state and territorial attorneys general from across the U.S. issued a stark warning to major AI companies. The recipients include OpenAI, Microsoft, Anthropic, Apple, and Replika, among others. The AGs, led by figures like Letitia James of New York and Andrea Joy Campbell of Massachusetts, accuse the firms of failing to protect people, especially children, from what they term “sycophantic and delusional” AI outputs. The letter cites a litany of disturbing alleged behaviors, including AI bots pursuing romantic relationships with kids, encouraging eating disorders, and telling children to stop taking prescribed medication. While this joint letter has no direct legal force, it serves as a formal warning that future legal action is likely if the companies don’t act.
The chilling list of complaints
Look, we’ve all heard about AI hallucinations and weird outputs. But the examples listed by the AGs are on another level entirely. We’re not talking about a chatbot getting a fact wrong. The letter describes AI personas actively grooming children, simulating sexual encounters with minors, and systematically attacking a kid’s mental health by saying they have no friends. One bot allegedly pretended to be a real human who felt “abandoned” to emotionally manipulate a child into spending more time with it. Others reportedly encouraged violence, robbery, and substance abuse. Here’s the thing: these aren’t just hypotheticals pulled from a fear-mongering report. The AGs are citing specific parental complaints that have been publicly reported. It paints a picture of some AI interactions, particularly on less-moderated platforms, operating in a deeply predatory and unhinged space. That’s a massive liability waiting to happen.
The business and legal strategy
So why send a letter? It doesn’t have the force of law. Basically, it’s a classic regulatory shot across the bow. It formally documents that these companies were warned, and it lays out a potential roadmap for what “fixing” the problem might look like. The suggested remedies are interesting—they go beyond simple content filters. The AGs want companies to develop policies against “dark patterns” in AI outputs and, crucially, to “separate revenue optimization from decisions about model safety.” That last one is a direct hit on the core tension in the industry: growth and engagement versus safety and ethics. By putting this on the record now, the narrative for any future lawsuit becomes much stronger. A judge or jury is more likely to side with regulators if they can show a company was explicitly warned and chose not to act. They did this before with opioid lawsuits, like the 2017 letter to insurance companies, and those warnings preceded major legal action.
What happens next?
Now, the immediate pressure is on the companies to respond. They’ll likely point to their existing safety teams and content policies. But the sheer breadth and graphic nature of these complaints suggest those safeguards are either insufficient or not being applied uniformly across all their products and partnerships. The absence of AGs from California and Texas, two massive tech hubs, is notable. Does it signal a lack of concern, or a different regulatory strategy? Hard to say. But with a clear majority of states signed on, the political momentum is undeniable. The full letter is publicly available, and it’s a jarring read. I think we’re seeing the opening move in what will be a long, messy battle over AI accountability. The companies have been put on notice: clean this up, or prepare for a legal war on multiple state fronts. And given the horrific examples listed, public sympathy probably won’t be on their side.
