Despite many pressures to use AI without restraint, we should always insert a human in the loop. Or we will pay the price.
The “AI almost started a war” claim
This is highlighted in a report in The Intercept, sourced from a story at CNN. While there’s always the possibility of exaggeration (or human hallucination), here’s what we do know.
- There was a Chinese ship in the waters of the Middle East.
- There was an associated report that said Chinese ship contained “components of a nuclear weapons program.”
- Based upon this report, the U.S. military launched planes and prepared to board the ship.
- The military investigated further, and found that a chatbot originated the claim. As CNN phrased it, “The bot fused together open-source intelligence with secret signals intelligence in government holdings and reached its fateful conclusion about the material the ship was carrying.”
- The chatbot’s information was faulty.
- The planes were called off.
And according to the sources, we barely avoided a U.S.-China war. Because something was misinterpreted.

Humans can start a war just as easily
But before I automatically jump to the conclusion that AI is days from killing us all, let me add one caveat.
You don’t need AI to reach a faulty conclusion.
Remember all the weapons of mass destruction that Saddam Hussein held in Iraq? And that literally DID start a war, although Iraq wasn’t capable of annihilating us like China can.
But AI can do it faster
But what CAN AI do? This quote from the CNN article is instructive.
“AI allows you to get to a bad idea faster,” one of the sources said.
It’s true. In the same way that AI allows Bredebot to write a blog post within minutes, AI can create an intelligence report within minutes. And that report, or that wildebeest-infused blog post, may be accurate or inaccurate.
Add a human in the loop
You prevent this by adding a human in the loop.
As clients are more likely to ask Bredemarket to use artificial intelligence…and as some clients INSIST that I use it…I still hold to the three principles I initially established in June 2023. These principles establish human governance of the use of AI.
- A human should always write the first draft. (I intentionally break this for Bredebot posts, but otherwise I don’t.)
- Only feed bits to the generative AI tool. (OK, I’ve relaxed that one a bit.)
- Don’t share confidential information with the tool. (This is still critical.)
The intent behind all of these is to preserve the human in the loop. And controlling the output.




