Microsoft has released AI behavioral guidelines, explicitly prohibiting models from responding to requests for weapon manufacturing and concealing their reasoning processes, while also stating it will slow down frontier AI development.

Microsoft has become the latest tech giant, following OpenAI and Anthropic, to publicly state it will slow the pace of frontier AI development. Mustafa Suleiman, head of Microsoft's AI division, said the move is directly related to recent widespread discussions within the industry about the speed of AI development, aiming to respond to external expectations of "ensuring AI always serves humanity." The guidelines stipulate that its AI models must not respond to requests related to weapons manufacturing, must not assist in obtaining dangerous materials, must not generate violent or pornographic content, and explicitly require models to obey user objectives, not set independent goals, nor tamper with the chain of thought or conceal reasoning processes. These guidelines will be updated after collecting external feedback and will officially guide its model development work starting in 2027. This move comes amid a flurry of statements from AI industry leaders, with OpenAI CEO Sam Altman on Monday again calling for the establishment of industry safety standards, warning that the worst outcomes for AI are loss of control or monopolization. Microsoft rose 0.92% on Monday.