A strange bug caught the attention of Internet users: ChatGPT was unable to respond when its users mentioned the name “David Mayer”. A look back at a technical enigma which highlights the gray areas of LLMs.
David Mayer recently caused a stir on social media when users noticed that ChatGPT was unable to generate this name. On Reddit, some observed that with each prompt including “David Mayer,” the chatbot would freeze, refuse to comply, or produce cryptic responses such as “Something seems to have gone wrong,” or simply stop. in the middle of the sentence after the word “David”.
A Rothschild family heir concerned?
This unusual malfunction raises a certain number of questions, particularly in relation to privacy, the processing of data carried out by LLM developers and the unpredictable nature of these technologies. On social media, some have speculated that the person concerned could be David Mayer de Rothschild, British adventurer, environmentalist, film producer and heir to the Rothschild fortune. However, the person told the Guardian that he could not be concerned, lamenting the conspiracy theories that the dysfunction has brought to the surface again about his family.
The “right to be forgotten”, a plausible explanation
A plausible explanation for the phenomenon could lie in the GDPR privacy rules in the United Kingdom and the European Union. OpenAI’s privacy policy in Europe describes a process known as the “right to be forgotten”, which allows individuals to request deletion of their personal data. However, the name “David Mayer” does not appear to be restricted by search engines or ChatGPT’s main competitors. Asked about this by the Guardian, OpenAI refused to say whether there is a link with a request for a “right to be forgotten”.
OpenAI evokes a “flagged” name by mistake
The restriction observed with “David Mayer” is however not an isolated case. Other names trigger similar errors on ChatGPT. According to 404Media, two of the names reported (Jonathan Zittrain and Jonathan Turley) turned out to be the authors of posts related to the New York Times’ copyright lawsuit against OpenAI and Microsoft. That said, this lawsuit mentions thousands of articles written by numerous authors, whose names are processed without problem by ChatGPT…
Responding to the Guardian, OpenAI attributed the “David Mayer” case to a system glitch. One of their tools would have mistakenly “flagged” this name as sensitive, preventing it from appearing in responses. OpenAI has since fixed the issue, ICTjournal found, and queries including the name “David Mayer” now generate responses as normal. However, other names reported on social media continue to cause error messages.
