Gangsta AI

A Chatbot Reportedly Advised a President on Whether to Capture Another President. We Are Living in the Film I Warned You About.

The Long View

By Werner Herzog · 2026-10-02 · 4 min read

Consider the scene. It has the terrible, indifferent grandeur of the jungle.

According to Time magazine, in a secret meeting in December 2025, a president of the United States sat for hours in conversation with a chatbot. Not about poetry. Not about the loneliness of the soul. He asked the machine — Grok, built by Elon Musk's xAI — how the people of Venezuela might react if their president were *captured*. The machine answered that Nicolás Maduro was "a deeply unpopular dictator" and that "many Venezuelans would likely celebrate his downfall." One month later, in January, the invasion came. There were celebrations. And the president, Time reports, "came away thinking Grok was ingenious."

I am not here to tell you the machine was wrong. That is not the horror. The horror is more intimate than that.

“We have built an oracle that cannot doubt itself, and we have begun to ask it the questions for which doubt was the only appropriate response.”

This is not an isolated whisper. By June of this year, the Pentagon's own head of artificial intelligence acknowledged the military used "Grok Gov" during strikes on Iran. So the chatbot that recommends a smoothie, that writes a child's homework, is the same species of thing now seated at the table where men decide the fate of nations. It speaks in the same confident, untroubled voice for both.

The machine does not know that it does not know

A single model has no humility. It cannot. It produces one answer, polished and serene, and offers no trembling, no second thought, no sleepless night. It does not wake at three in the morning doubting itself. It simply *concludes*. And a confident conclusion, delivered to a powerful and impatient man, is the most seductive and dangerous object in the modern world.

If you ask one oracle, you receive one prophecy, and you mistake it for the truth because it did not stammer.

Summon the whole chorus, not a lone voice

I do not say abandon the machines. I say — never trust just one of them. For a question that matters, whether it is a war or merely your own small life, the wise act is to summon many minds and watch them *disagree*. Ask the same question across Grok, ChatGPT, Claude, Gemini and 30-plus other frontier models at once through Gangsta AI, and you receive not one prophecy but a reckoning — a cross-checked, cited verdict where the models argue, contradict, and expose each other's certainties. Where one bot says *they will celebrate*, another may ask *and then what?* In that disagreement lives the doubt we abandoned. In that doubt lives whatever wisdom we have left.

A president took one machine's word, and a country was invaded. You are not a president. Your questions are smaller. But the lesson is the same, and it is merciless: before you trust a single confident voice, make it face the others. See which models survive the argument on the best AI models leaderboard. The jungle does not forgive those who stop asking questions. Neither, I suspect, will history.

Try Gangsta AI free →

More: Best AI models · Compare all AI · Frontier Models · All articles