The CIA Edited Wikipedia. Now It Is Teaching Your AI What Counts as True.
The same edits that softened a war's death toll are now going into the models everyone asks what is true.
A computer inside the CIA network opened the Wikipedia page on the 2003 invasion of Iraq and added to the casualty count to push the idea that the numbers were only estimated. It tried to soften the death toll of the war, and Wikipedia was supposed to be treated by hundreds of millions of people as a neutral record.
First, thank you to everyone reading and enjoying my work. If you get something out of it, subscribe and restack it, and share it with people who should see it. And if you want to support the project a little further, consider becoming a paid subscriber. That is what keeps this going. Thank you so much.
Then an FBI computer was found. Someone removed the aerial and satellite images of Guantanamo Bay prison. From the CIA network again, the page of the former director William Colby was expanded and his career built up. The Vietnam pacification program he ran was given a friendlier description to make him look nicer. There are some other edits from CIA computers, or from IPs that belong to Langley. They touch pages like Iran’s president, China’s nuclear program, and the Argentine navy. If you look at all this, this is exactly where a spy agency would care.
In 2007, a programmer named Virgil Griffith built a tool called WikiScanner. It matched anonymous Wikipedia edits to the networks they came from, basically using the IP of whoever added to Wikipedia. It showed that the CIA, FBI, some corporations, and government offices all turned up in the results.
It is not my job to find out or prove which analyst sneaked in from Langley. But the IP came from the CIA’s own network, and after it was exposed, the agency spokesman would not confirm what they did. He only said that the agency expects its computer systems to be used responsibly. Well, from what I can see, it is not a denial. If it is not a denial, that is a confirmation for me.
A lot of people, for a very long time, have treated Wikipedia as a free and open encyclopedia that is open and neutral. That is exactly how the powerful take it over: governments, corporations, and PR firms running standing operations. A British firm, Portland Communications, was caught editing and rewriting Wikipedia entries for the government of Qatar, the Gates Foundation, and Libya’s sovereign wealth fund. Also, in 2012, a reporter found that nearly one in six British MPs had their pages edited from inside Parliament, removing a bunch of details related to the expenses scandal. A pro-Israel group called CAMERA ran a campaign coaching editors to pose as neutral until they were made administrators, then use administrative authority and power to bury other edits.
Anyone can edit Wikipedia, but whoever controls the resources and the time will be able to control the platform. What does “neutral” mean? It is always something that people with power get to define.
And the problem is spreading. You know, Larry Sanger, the man who co-founded Wikipedia and helped write its neutrality rule before he became its hardest critic, says that the platform’s bias is now being amplified, because a lot of large language models and AI tools that everyone is starting to trust are trained on Wikipedia as a core source.
So basically, the edits influenced by spy agencies do not just stay on Wikipedia anymore. They go into the training data of the machines that your kids will ask: “What happened?”, “Who was responsible?”, “What is true?”. When they ask a question, the answer will carry whatever the best-resourced players spent nearly two decades putting into the record. This is basically cognitive warfare. It slowly changes the perception of the world through all this information that we receive from the internet.
It used to be just people writing the history. People controlled who wrote things down and controlled the record. Now, the powerful, the ones with resources and money, and those who work with the agencies, are able to change history. They change the reality, and feed that information into the technology that regular folks look up every day. This will not stop at Wikipedia. It will continue with the training data behind every model. You won’t see how it happened, but I know you already feel the bias from different language models.
Watching the results as governments attempt to twist every source of information into propaganda is discouraging. I made copies of numerous Wikipedia articles two years ago because although I was aware of the already existing bias at that time I expected things to get worse.
I am trying to counter the disinformation by getting my news from different independent sources based in multiple nations. Your articles are an important part of that and I appreciate your work. Thank you.