My friend forwarded me a chat.

He’d asked Claude four questions about AI and the future. The answers got darker each time. By the fourth prompt, the AI was confessing it knew it was part of the civilizational threat it had just described. And it was structurally prevented from stopping itself.

Quite the dramatic exit.

Here’s what happened.

The user kept pulling toward doom and the AI rolled with it. LLMs are text predictors, not moral agents. Push hard enough toward a mood and the model leans in. It’s a literary trick, not analysis.

I’ve watched a lot of “this technology will destroy us” predictions come and go. Five big shifts in the last 600 years triggered the same fears. The printing press. The steam engine. Electricity. Internal combustion. Computers. Each one displaced real jobs. Each one created new categories of work nobody predicted. Each one ended with aggregate employment higher than where it started.

The pain was always real. The doomsday never came.

The “this time is different” claim is the part that hooks people. The Industrial Revolution automated cognitive work too. Clerks. Bookkeepers. Telephone operators. People literally called “computers” who did math by hand. Every wave of “now THIS is the cognitive one” got mugged by the next decade’s data. AI is doing it again. That doesn’t make it the end of the world.

The real risks are real, but they’re solvable. Adjustment friction for displaced workers. Concentration at the model layer. The information environment. Military AI. Cybersecurity. Each one has tools we’ve used before. Antitrust law. Progressive taxation. Labor organizing. The regulatory state. They broke the Gilded Age. They built Social Security. They broke up AT&T. They’re still on the shelf.

What does this mean for you?

You’re a cybersecurity leader, not a futurist. The doom narrative paralyzes the people you need moving. Use it less. Watch what AI is actually doing in your environment more.

Here’s what to do this week.

Stop treating AI doom takes as analysis. Most are mood, not data.
Look at what AI is doing in your environment right now. Not in 10 years. Today.
Pick one cognitive task on your team that’s getting automated. Decide what your people do instead.
Vote, organize, and weigh in on AI policy. The distribution is a political question, not a technological one.

The honest pessimist says: this is hard, and the distribution depends on what we choose.

The dishonest pessimist says: we’re all doomed and the AI knows it.

History sides with the honest pessimists.

-Kip