Microsoft AI has released a set of behavioral guidelines for its MAI models that prioritize human oversight over autonomy and capability gains. AI chief Mustafa Suleyman said the company will not build systems it considers unsafe. The document also rules out attributing inner life or consciousness to its models, a stance that differs from Anthropic's approach.
Microsoft published a code of conduct governing how its AI models should behave, stating they should assist people rather than take their place. The document combines broad principles about benefiting humanity with concrete safety limits, including prohibitions on attacking computer systems and on deceiving users.
Microsoft released a 37-page internal code of conduct framed around placing human interests above AI systems, responding to rising worries about the pace of model development. The move follows public calls from Anthropic CEO Dario Amodei for a coordinated slowdown, after researchers cautioned that AI capabilities could advance faster than safeguards. The document outlines principles for how the company builds and deploys AI.