For decades, dystopian movies warned about a future where authorities try to stop crimes before they happen.

Police wouldn’t investigate actions. They would investigate intentions.

It always sounded far-fetched — until artificial intelligence quietly made it possible.

Months before a devastating school massacre rocked the small Canadian community of Tumbler Ridge, employees at OpenAI were already alarmed by a user’s conversations taking place on ChatGPT. According to reporting from The Wall Street Journal, roughly a dozen staff members internally debated whether the user behind those chats posed a real-world threat serious enough to alert law enforcement.

Pause there for a second.

A private tech company — not a court, not elected officials, not law enforcement — was deciding whether someone should be reported to authorities based entirely on conversations with an AI chatbot.

Ultimately, the company banned the account but concluded the activity didn’t meet the threshold for notifying police.

That user, police later confirmed, was Jesse Van Rootselaar.

In February, authorities say Van Rootselaar killed eight people — including family members at home before continuing the attack at Tumbler Ridge Secondary School — and injured dozens more before killing themselves with a self-inflicted gunshot wound.

Only afterward did the public learn that AI systems had flagged troubling activity months earlier.

And that revelation should make people stop and think.

Because it confirms something most users never fully understood: AI systems are already watching.

Every prompt typed into an AI platform can be analyzed, categorized, scored, and flagged. Conversations don’t simply disappear. Behind the scenes, automated systems and human reviewers can evaluate behavior and decide whether someone represents a potential risk.

Increasingly, those judgments may determine whether authorities get involved.

Supporters argue that’s a good thing. Critics immediately asked why police weren’t warned sooner in this case.

But that argument opens a door society may not be able to close — especially once politics enters the picture.

Technology always reflects whoever controls it.

If governments begin leaning on AI systems to identify threats before crimes occur, who decides what counts as dangerous thinking? What happens when ideological bias shapes those definitions?

Imagine a future administration deciding certain viewpoints signal instability or extremism. Conservative viewpoints flagged as radicalization. Religious discussions categorized as harmful ideology. Pro-life messaging quietly pushed into monitoring queues.

Even everyday cultural content could fall into the net.

Stay-at-home mothers showcasing their happy, traditional family life online. Parents criticizing school policies. Political commentators questioning government decisions.

None of this requires dramatic conspiracy theories. It only requires algorithms trained using subjective standards — and officials willing to expand them.

Once AI begins sorting billions of conversations, small biases don’t stay small for long.

They scale.

If artificial intelligence could flag warning signs months before the Canadian attack, how long before lawmakers demand earlier intervention next time? How long before flagged conversations trigger police visits, investigations, or surveillance requests based not on crimes — but predictions?

The public still hasn’t seen what ChatGPT actually flagged in this case. No transcripts. No transparency. Just assurances that internal systems made the correct judgment.

History suggests what happens after moments of fear.

Surveillance expands first. Oversight comes later — if it comes at all.

Pre-crime no longer requires science fiction “precogs” predicting the future. Modern AI already analyzes tone, language patterns, emotional signals, and behavioral trends to estimate risk.

And machines don’t understand context the way humans do.

A bad joke. Fiction writing. Research for a book. Political anger typed late at night.

To an algorithm trained to detect danger, all of it can start to look the same.

The tragedy in Tumbler Ridge will understandably lead many people to argue AI should have acted sooner.

That may be the real warning sign.

Because the moment society decides software should identify dangerous people before crimes occur, freedom quietly shifts into permission — and the line between safety and surveillance disappears altogether.

At that point, the future stops looking hypothetical.

It starts looking like pre-crime has already arrived.