A Russia-linked network uses AI to rewrite real news stories

A network of websites known as CopyCop has been accused of publishing deceptive posts, with 19,000 instances recorded in a single month. The posts, written in English and French, were taken from legitimate news outlets and altered using large language models, likely one built by OpenAI. The stories covered a range of contentious issues, including accusations of war crimes against Israel, amplification of divisive political debates in America, and spreading false stories about Polish mercenaries in Ukraine. The articles were translated and edited to add a partisan bias, with some even showing visible instructions to the AI model to rewrite the articles with a specific political stance.

The method used by CopyCop mirrors the disinformation tactics used by the KGB in the 1980s, but technology has accelerated the process. An investigation by Recorded Future, a threat-intelligence company, revealed that the articles had been manipulated to promote a particular viewpoint. This marks a new development in Russian propaganda, where AI technology is used to modify legitimate news stories to spread disinformation and fuel political divisions.

Key takeaways

A network of websites called CopyCop has been publishing modified stories from legitimate news outlets, using large language models to add a partisan bias.
The stories covered a range of contentious issues and were published in both English and French.
It is suspected that the language model used for this operation was built by OpenAI, the American firm that operates ChatGPT.
A threat-intelligence company, Recorded Future, published an investigation on May 9th revealing that the articles had been translated and edited to add a partisan bias.

A Russia-linked network uses AI to rewrite real news stories

Key takeaways

Discussion (0)