Lies Our Machines Tell Us: Why the New Generation of “Reasoning” AIs Can’t be Trusted

A flood of advanced new artificial intelligence models is upon us, led by China’s DeepSeek. They purport to “think” and even to explain their reasoning. But are they really a step forward? In this original investigation, Gleb Lisikh – who previously took on ChatGPT to probe its political biases – engages with DeepSeek in a debate about systemic racism. Lisikh finds it doesn’t just spout propaganda but attempts to convince him using logical fallacies and outright fabrications. In a future where virtually all information and communication will be digital, a dominant technology that doesn’t care about the objectivity and quality of the information it provides – and even actively misleads people – is a terrifying prospect.