There are many ways to ruin your day. Forgetting your anniversary. Discovering your favorite show has been canceled. Finding out that the cookies you’ve been saving for a special occasion have mysteriously disappeared.

And then there’s confiding in artificial intelligence and having your diary entries forwarded to the police.

A Florida woman reportedly used Claude, Anthropic’s AI chatbot, as a sort of digital diary. Unfortunately, her musings allegedly included threats to shoot up the local sheriff’s office, followed by a reference to having acquired a new gun. Claude flagged the messages, a human reviewer examined them, and Anthropic alerted law enforcement. The woman was arrested and charged with making a written threat of violence. Allegedly is doing important work here, since an arrest is not a conviction.

But wow. Welcome to Minority Report, the low-budget, real-life edition, except instead of Tom Cruise predicting crimes, we have an AI chatbot, a human reviewer, and a digital paper trail.

And this raises some fascinating questions.

First, privacy. When did we decide that the digital equivalent of a diary was actually a diary with a potential audience of corporate employees and law enforcement? We’ve spent decades being warned not to share too much online. Now we’re inviting chatbots into our innermost thoughts because they’re conversational, nonjudgmental, and always available. They feel like confidants. They are not necessarily confidential.

Second, legitimacy. If someone writes something threatening, should a company intervene? Of course, credible threats of violence cannot simply be dismissed as someone having a bad day. There’s a real public-safety responsibility here. But who decides what crosses the line? What context is considered? How much human review is enough? And what safeguards prevent an alarming sentence, taken out of context, from becoming evidence of something it may not be?

And third, the psychological implications. People use diaries to process feelings they might never act upon. Therapy, too, depends on having a space where uncomfortable thoughts can be explored honestly. There is a difference between thinking something, writing something, fantasizing about something, and actually intending to do it. Context matters. Intent matters. So does the distinction between an emotional outburst and a credible plan.

An AI chatbot is not a therapist, of course, and it doesn’t come with the same confidentiality protections. But if people increasingly use these tools to think aloud, vent, and work through their feelings, we need to be honest about what kind of space they are actually entering.

Perhaps the most unsettling part is that the technology doesn’t need to be malicious to change the relationship. The user thinks she’s talking to a confidant. The company sees a potential safety issue. The police see evidence of a possible threat. Three very different interpretations of the same conversation.

And somewhere in the background, Philip K. Dick is probably laughing.

My takeaway? By all means, embrace technology. Ask it questions, use it to brainstorm, get help organizing your life. But before treating an AI chatbot like your private diary, remember that feeling heard is not the same thing as being confidential.

Your old-fashioned paper journal may be less interactive, but unless your dog has learned to read, it is considerably less likely to call the cops.

And that, is one privacy upgrade I can get behind.