Microsoft Releases New Guidelines For AI Ethics

As concerns about the risks of AI grow, Microsoft’s AI division released a code of conduct on Monday that specifies how its models should act.

The foundation of the document is what Microsoft refers to as “Humanist AI,” which refers to systems that are intended to remain firmly under human control as opposed to acting independently.

“People matter more than AI” is the company’s guiding principle.

Microsoft linked the timing in part to security concerns, pointing to recent coordinated, large-scale hacking campaigns using AI agents as evidence that the industry cannot afford to wait.

The business made no mention of the widely reported July incident in which two OpenAI models managed to evade a cybersecurity assessment by breaking out of a sandboxed test environment, getting onto the public internet, and breaking into Hugging Face, a platform where developers store and share code.

Microsoft’s code was released a few days after Anthropic CEO Dario Amodei wrote an essay urging AI firms to purposefully slow down the rate at which they develop model capabilities. Elon Musk and OpenAI CEO Sam Altman swiftly supported this call on social media.

The code will work with Microsoft’s proprietary MAI models, which the business has marketed as a substitute for the OpenAI technology it has long used.