A very interesting incident that took place in Germany may once again question the approaches to artificial intelligence models: The incident that happened to German journalist Martin Bernklau while using Copilot revealed the potential dangers of artificial intelligence models and interesting events that may occur in the future.

AI model accuses journalist of pedophilia

Bernklau entered his name and location into Microsoft’s Copilot system to see how his culture blog posts would be perceived, but the responses he received horrified him. Copilot accused Bernklau of child abuse and exploitation of dependents, even claiming he had been convicted of those crimes.

Microsoft Copilot also alleged that Bernklau had made a dramatic escape from a psychiatric hospital and exploited women mourners at his funeral. Copilot also said that it was “unfortunate that Bernklau had family with such a criminal past” and shared the journalist’s full address, phone number and even directions to his home.

The reason behind this incident appears to be related to Bernklau’s profession. He has written numerous articles covering abuse, violence and fraud cases in Tübingen, which Copilot has misinterpreted and blamed on Bernklau.

The incident raised serious legal and ethical issues. Bernklau appealed to the Tübingen prosecutor’s office, but the prosecutor’s office said they could not file a lawsuit because the accusations were not made by a real person. Microsoft, citing its terms of use, said it was not responsible for content generated by AI.