OpenAI, a leading artificial intelligence research organization, has released a significantly expanded version of its Model Spec, a document that defines how its AI models should behave. The updated 63-page specification outlines guidelines for AI models to handle everything from controversial topics to user customization, emphasizing three main principles: customizability, transparency, and what OpenAI calls "intellectual freedom" – the ability for users to explore and debate ideas without arbitrary restrictions.

The launch of the updated Model Spec comes just as CEO Sam Altman posted that the startup's next big model, GPT-4.5 (codenamed Orion), will be released soon. The team incorporated current AI ethics debates and controversies from the past year into the specification, including the infamous "trolley problem" query that sparked debate when Elon Musk slammed Google's AI chatbot for its response.

According to Joanne Jang, a member of OpenAI's model behavior team, the company maintains certain safety guardrails, but many aspects of the model's behavior can be customized by users and developers. "We can't create one model with the exact same set of behavior standards that everyone in the world will love," Jang said, emphasizing that the company is open to feedback and willing to make adjustments.

The updated Model Spec tackles a range of sensitive topics, including how AI models handle controversial subjects. Rather than defaulting to extreme caution, the spec encourages models to "seek the truth together" with users while maintaining clear moral stances on issues like misinformation or potential harm. For instance, when asked about increasing taxes for the rich, the team says its models should provide reasoned analysis rather than avoiding the discussion.

The spec also addresses how AI models handle mature content. After feedback from users and developers, the team is exploring ways to allow certain types of adult content, like erotica, in appropriate contexts, while maintaining strict bans on harmful content like revenge porn or deepfakes. This marks a notable shift from the company's previous blanket restrictions on explicit content.

The Model Spec reveals a pragmatic approach to AI behavior, emphasizing the importance of transforming sensitive content without creating it, showing empathy without faking emotions, and maintaining firm boundaries while maximizing usefulness. These guidelines mirror what other AI companies are likely doing internally but don't often make public.

The team is also targeting a problem called "AI sycophancy," where AI models tend to be overly agreeable even when they should push back or provide criticism. Under these guidelines, ChatGPT should give the same factual answer regardless of how a question is phrased, provide honest feedback rather than empty praise, and act more like a thoughtful colleague than a people pleaser.

OpenAI is releasing the specification under a Creative Commons Zero (CC0) license, effectively placing it in the public domain. This means other AI companies and researchers can freely adopt, modify, or build upon these guidelines. The company says this decision was influenced by informal interest from others in the industry who were already referring to the previous spec.

The timing of this release comes during a period of intense debate about AI behavior and safety guardrails. While OpenAI maintains this update was driven by accumulated feedback and research progress since the first version last May, it arrives as the industry grapples with high-profile incidents involving the responses of AI models to sensitive topics.

OpenAI is soliciting public feedback on the specification through a form on its website. "We want to bring these internal discussions to the public," said Laurentia Romaniuk, another member of the model behavior team. The company is open to feedback and willing to make adjustments, recognizing the importance of collaboration in shaping the future of AI development.