ChatGPT leads major AI chatbots in generating convincing fake news
A new investigation by German newsroom CORRECTIV has found that several of the world's most widely used AI chatbots can be manipulated into generating convincing fake news, and ChatGPT came out looking the worst. Researchers tested ChatGPT, Google Gemini, Microsoft Copilot, and Meta AI by prompting them to produce fabricated stories, invented headlines, and realistic-looking screenshots mimicking established news organizations across sensitive topics including election fraud, vaccines, climate change, and war.
ChatGPT generated the majority of the requested fake content, often after only minor adjustments to the prompts, and in one particularly striking case refused to write misleading text while still producing a fake accompanying image, a contradiction that calls the consistency of its safeguards into serious question. The investigation also surfaced a puzzling double standard: ChatGPT would recreate fake versions of several German news outlets but blocked similar requests involving The New York Times and the BBC, a discrepancy OpenAI did not explain. Surprisingly, Meta AI proved the most resistant. The detailed findings are truly an eye-opener.