AI Models Need to be Disinfected — Or George Orwell’s “1984” Will Come True

NewsGuard By Gordon Crovitz, NewsGuard Co-CEO: “Recent forecasts say the AI companies could soon spend trillions of dollars to expand their AI offerings. In addition to paying for expensive chips and giant server farms, the AI companies should make their AI output trustworthy. They have allowed their models to be infected with malign falsehoods so deeply that they deliver on George Orwell’s prediction in “1984” that falsehoods would become perceived as truth. That is the lesson from a sobering analysis published last week in an academic journal of how Vladimir Putin’s government has undermined the AI models by getting them to spew the Kremlin’s favorite false claims. Huw Dylan and Elena Grossfeld of the Department of War Studies at King’s College London entitled their paper, “Revisionist future: Russia’s assault on large language models, the distortion of collective memory, and the politics of eternity,” published by the journal Dialogues on Digital Society.

The authors peg their concerns to the article “The AI Disinformation Battlefield” in the technology journal Enterprise Security Tech that details the results of a NewsGuard assessment. Our report issued in March showed that a Russian campaign called the Pravda Network had infected the AI models so badly that on average the models repeat false Kremlin claims one third of the time when people ask the chatbots about them. Using AI as a force multiplier to spread its disinformation at breathtaking scale, this network spread the Kremlin’s 207 favorite false claims by concocting and spreading 3.6 million articles in 2024 alone, using 150 fake news websites targeting 49 countries in dozens of languages, NewsGuard found. In other words, the Russians used AI to infect AI globally. NewsGuard analysts tested the 10 largest AI models and found they routinely spread these Pravda Network false claims. Among the fabricated claims the models spread was that Ukrainian President Volodymyr Zelensky blocked U.S. President Donald Trump’s Truth Social app in Ukraine, that a Ukrainian battalion burned an effigy of Trump, and that Zelensky spent 14.2 million euros in Western military aid to buy the Eagle’s Nest estate frequented by Adolf Hitler. Dylan and Grossfeld write that this means “Users consulting LLMs are increasingly receiving subtly contaminated responses: authoritative seeming answers imbued with conspiracy theories, fabricated narratives, or questionable interpretations.”

Posted in: AI, Censorship, Civil Liberties, Internet, Knowledge Management, Legal Research, Search Engines, Social Media