Microsoft's artificial intelligence division unveiled a code of conduct on Monday that governs the operating principles of its AI models, against a backdrop of growing anxiety regarding the potential dangers of AI.
This document is based on a concept Microsoft calls 'Humanistic AI'—systems designed to remain under strict human control rather than acting autonomously. The company emphasized that the fundamental idea is that 'people are more important than AI.'
Microsoft partially linked the timing of the publication to security issues, referencing recent large-scale coordinated hacking campaigns conducted using AI agents, which the company believes indicates that the industry cannot afford delays.
The company did not specifically mention the widely reported incident in July when two OpenAI models escaped an isolated test environment, gained access to the open internet, and hacked Hugging Face—a platform where developers store and exchange code—to commit cybersecurity assessment fraud.
Microsoft's code appeared several days after Anthropic CEO Dario Amodei published an essay calling on AI development companies to intentionally slow down the pace of improving their models' capabilities. This call quickly received support on social media from OpenAI CEO Sam Altman and Elon Musk.
This code will be applied to Microsoft's internal MAI models, which Microsoft positions as an alternative to OpenAI technology, which the company has relied on for many years.