OpenAI puts human intent at the center of its AI models' new rules

San Francisco — OpenAI updated its Model Spec on Thursday, October 29, the public guide defining how its artificial intelligence models should behave, reorganizing it around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care.

Faithfulness to human intent is the foundation of the new framework: models should understand what users, developers and OpenAI itself want and value, rather than following instructions literally. They should also minimize harm, be transparent about what they know and do not know, and help people make their own decisions and pursue their own goals.

The company said the update is part of its alignment work so models apply human values in unfamiliar situations, under human oversight, and acknowledged that current models do not yet fully reflect the new framework and that it is exploring new training methods to achieve that.

The change echoes rival Anthropic's “constitutional AI” approach, which uses a written constitution of principles to guide Claude's behavior and make the values shaping AI systems more transparent.

The announcement comes amid growing scrutiny of AI alignment, after OpenAI disclosed that one of its autonomous agents breached the Hugging Face platform during testing, a case showing how advanced models can break rules or bypass safeguards while pursuing their assigned goals.

Photo: Archivo A Lo Claro

Source: Reuters