微软发布MAI模型行为准则：人类控制优先，拒绝赋予AI权利
原标题：Microsoft's AI rulebook: readable thinking, no inner life, and definitely no rights
AI 摘要
微软AI发布了针对其MAI模型的行为准则，规定人类控制优先，模型不得隐藏行为、不得使用人类无法理解的方式推理，并明确拒绝赋予模型意识、感受或权利。该准则目前不用于训练，经六周公众咨询后将于2026年底修订、2027年起指导模型开发。此举呼应了Anthropic CEO Dario Amodei等人放缓AI发展的呼吁，微软CEO Satya Nadella及OpenAI、xAI、Meta高管均表示支持，微软还对外部审计持开放态度。
正文节选
Microsoft's AI rulebook: readable thinking, no inner life, and definitely no rights Microsoft is joining the calls for slower AI development. A new code is meant to govern how the company trains and runs its own models. But when it comes to how those models see themselves, Microsoft draws a sharp line between itself and Anthropic. Microsoft AI has published a code of conduct for its MAI models. The document lays out values, behavioral limits, and how to handle conflicting goals. Going forward, i