NE TAPEZ JAMAIS CE NOM SUR CHATGPT (ou il va sauter !)

NE TAPEZ JAMAIS CE NOM SUR CHATGPT (ou il va sauter !)

by Ludo Salenne

Applied Sciences & Engineering Computing & Cybersecurity UBDigital and Information technology: general topicsUBWInternet: general works
🎙 Ludo Salenne 👥 267K 📅 December 2, 2024 ⏱ 20 min 👁 17K 📄 documentary 🧭 2026-08-21
Available in: English (current) Français

Keywords

David MayerChatGPT bugOpenAIprompt injectionstop sequences

Summary

The video investigates a peculiar bug in ChatGPT where the name ‘David Mayer’ triggers an error, preventing the AI from generating responses. The creator tests various models (GPT-4, GPT-4o, o1-preview) and attempts to bypass the restriction using different prompts, including a ‘developer mode’ prompt and a cryptographic substitution technique. He also explores the possibility that the issue is related to a specific individual, possibly David Mayer de Rothschild, or a privacy request under GDPR. The video includes a technical explanation involving ‘stop sequences’ and discusses various theories circulating online. The creator also promotes his training course and asks viewers whether they would want their personal information removed from the internet. The investigation is hands-on but lacks conclusive evidence, leaving the root cause unresolved.

123 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides practical demonstrations of the bug and several workarounds, which is valuable for users encountering similar issues. The argumentation is structured as an investigation, with the creator testing hypotheses and presenting evidence (e.g., screenshots, LinkedIn profiles). However, the reasoning often relies on speculation and anecdotal evidence rather than solid data. The creator acknowledges the lack of confirmation and invites viewers to share their own theories, which adds a collaborative element but also weakens the scientific rigor.

Scientific Rigor, Source Quality, Title Accuracy

The video cites no external sources beyond the creator’s own resources and tutorials. The investigation uses publicly available information (LinkedIn, Wikipedia) but does not provide direct links. The title is sensationalist but matches the content’s focus on the bug. The creator does not clearly distinguish between verified facts and hypotheses, which could mislead viewers. The video’s scientific rigor is limited by the absence of authoritative references and the speculative nature of the conclusions.

166 words

Title / Content Match

The title is clickbait but accurately reflects the video's focus on the David Mayer bug and attempts to bypass it.

Quality & Reliability

6/10

The video presents a mix of hands-on testing and speculative theories about a ChatGPT bug. The creator demonstrates several workarounds and provides technical explanations (e.g., stop sequences), but the investigation relies on unverified assumptions and lacks authoritative sources. The content is engaging but not rigorously scientific.

Chapters

Cited Sources

Concurring Sources

  • OpenAI API documentation on stop sequences — Supports the technical explanation of stop sequences as a possible cause.
  • GDPR Article 17 - Right to erasure — Supports the theory that a privacy request could lead to name blocking.

Dissenting Sources

  • OpenAI official statement on the bug — No official statement was found; the video speculates without confirmation.

Contribution & Novelties

The video offers a practical, hands-on exploration of a specific ChatGPT bug, demonstrating various bypass techniques and discussing potential causes. It contributes to the community’s understanding of such issues, though it does not provide a definitive answer.

Pour aller plus loin :

  • Stop sequences in language models — Official documentation explaining how stop sequences work, relevant to the technical explanation.
  • GDPR and the right to erasure — Legal basis for individuals requesting removal of personal data, a theory discussed in the video.
  • David Mayer de Rothschild — Wikipedia page of the environmentalist, one of the speculated identities.
  • Prompt injection attacks — Overview of prompt injection, a related security concept.

109 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with a slight peak in information quantity and a dip in reliability. This reflects a video that is informative and technically engaging but lacks authoritative backing and rigorous verification.

Reliability 5/10

💬 Positive: The comments are largely positive, with viewers sharing their own tests and theories, and expressing appreciation for the video's content. Some mention the bug being fixed, while others engage with the privacy question posed by the creator.