Free daily briefing on global business news.
The guidelines bar Microsoft's models from helping manufacture weapons or concealing their reasoning, with a final version set to guide development starting in 2027
Bloomberg / Getty Images
Microsoft $MSFT posted a provisional code of conduct Monday that places restrictions on its in-house AI models, days after the leaders of Anthropic and OpenAI called for slowing the pace of AI development.
Under the new code of conduct, Microsoft's AI models — a family of systems the company brands as MAI — are barred from helping users with weapons manufacturing, facilitating access to hazardous materials, promoting disordered eating, or generating violent or sexually explicit content, according to CNBC. The models are further required to remain oriented toward what users actually want, rather than pursuing independent objectives or hiding what they are doing.
