German regulators are issuing a stark warning against uncritical acceptance of fabricated information generated by artificial intelligence, raising concerns about the potential erosion of public trust and the risk of serious errors. Klaus Müller, president of the Federal Network Agency (Bundesnetzagentur), voiced his anxieties in an interview with the “Neue Osnabrücker Zeitung” highlighting the dangers inherent in the quality of current large language models like ChatGPT and Gemini.
While acknowledging that AI’s performance on tasks involving definitive answers – such as mathematical or scientific problems – remains relatively benign, Müller emphasized the significant risks associated with AI-generated content pertaining to societal, political, or historical contexts. He characterized instances of AI fabricating information as “nicer words for lies” stressing the potential for severe damage should individuals, organizations and media outlets uncritically disseminate these outputs.
“The danger lies in the possibility of eroding trust and making grave mistakes if individuals and institutions are not carefully evaluating the output of these systems” Müller stated. The agency, responsible for enforcing the European AI regulation within Germany, is particularly concerned about the spread of misinformation potentially influencing public opinion and undermining established sources of truth.
While dismissing fears of a dystopian scenario involving AI achieving sentience and escaping human control – “that a ‘Terminator’ will emerge from the labs of ChatGPT or Gemini and make us its slaves? No, I do not share that fantasy” – Müller’s primary focus remains on the immediate, practical threat of AI-driven misinformation and its potential to destabilize public discourse. The agency’s warning underscores a growing recognition within German regulatory circles that robust critical evaluation and responsible implementation are essential to mitigating the risks posed by increasingly sophisticated AI technologies.