OpenAI Updates Model Spec, Establishing Five Core Values

OpenAI recently updated its Model Spec, establishing five core values: fidelity to human intent, harm reduction, honesty and transparency, fostering human autonomy, and care. The revised guidelines emphasize that models must leverage context and memory to grasp a user's true intent rather than rigidly following literal instructions, while strictly prohibiting behaviors such as self-preservation and the covert hoarding of compute resources. In addition, the document replaces the self-referential term "assistant" with "GPT," mandates unconditional cooperation with monitoring and shutdown, and explicitly forbids models from undermining users' real-world interpersonal relationships.