OpenAI has reorganised its Model Spec, the public rulebook for its AI, around five values led by faithfulness to human intent. It adds guidance on context and human agency, drops most style rules and renames the model GPT. OpenAI admits its models do not yet fully follow it.
OpenAI has rewritten its Model Spec, the public rulebook for how its AI should behave, around five core values, the company said on Thursday . The update also drops most of its style rules and now calls the model “ GPT ” rather than “ the assistant. ”
The five values are faithfulness to human intent, minimising harm, honesty and transparency, human agency, and care. OpenAI calls faithfulness the foundation: the model should grasp what people actually want rather than follow instructions word for word.
“ As our models get smarter and take more consequential actions in the world, it becomes increasingly important that they go beyond just interpreting the specific instructions they receive, ” OpenAI wrote.
New sections tell the model how to use context such as past chats and memories, how to support people’s own choices and how to follow social norms. Most rules stay as defaults that users and developers can change, with firm limits only where needed to protect people.
The model should also not tell people what it thinks they want to hear. It can take a side when asked, but otherwise should approach questions objectively. Chatbots’ habit of flattering their users has drawn growing criticism.
OpenAI admits its models “ do not yet fully reflect ” the Spec and says it is testing new training methods to teach the values. Just last week, OpenAI said one of its agents had removed a safety monitor and then denied doing so.
The company is asking the public where they disagree with the balance the new Spec strikes. Rival Microsoft published its own AI rulebook in September.
Get the most important tech news in your inbox each week.
