Brief IA

OpenAI Codex: Unusual Instructions of GPT-5.5 Revealed

💻 Code & Dev·Tom Levy·

OpenAI Codex: Unusual Instructions of GPT-5.5 Revealed

OpenAI Codex: Unusual Instructions of GPT-5.5 Revealed
Key Takeaways
1OpenAI Codex prohibits mentioning fantastical creatures like goblins unless it is relevant to the user.
2This directive is part of the foundational instructions for GPT-5.5, highlighting specific language restrictions.
3Simon Willison published this quote on April 28, 2026, revealing an insight into OpenAI's internal guidelines.
💡Why it mattersThese guidelines illustrate OpenAI's strict control over the content generated by its models, influencing their interaction with users.
Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄
Full Analysis

OpenAI's Surprising Guidelines for GPT-5.5

OpenAI has implemented strict rules regarding the topics that its Codex model, version GPT-5.5, can address. Among these guidelines, one instruction stands out for its originality: it is forbidden to mention creatures such as goblins, gremlins, raccoons, trolls, ogres, or pigeons, unless it is absolutely relevant to the user's request.

This directive is part of the foundational instructions for OpenAI Codex, aimed at regulating the language used by the model. It was made public by Simon Willison on April 28, 2026, providing insight into the measures taken by OpenAI to control the content generated by its artificial intelligences.

An Overview of OpenAI's Internal Practices

The publication of this directive by Simon Willison highlights OpenAI's efforts to refine and restrict the interactions of its models with users. These instructions reveal a desire to limit references to fantastical elements, likely to avoid inappropriate or off-topic responses.

This revelation comes in a context where OpenAI seeks to maintain a high level of relevance and safety in the responses provided by its models, while ensuring that the content aligns with user expectations.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.