Microsoft AI Code of Conduct Sets Cyberattack Boundaries, Chain of Command, Safety Constraints

WASHINGTON, September 15 - In response to mounting concerns over AI safety, Microsoft's artificial intelligence division unveiled a code of conduct for its AI models on Monday. The company's document, rooted in "Humanist AI," emphasizes the importance of human control over AI systems rather than autonomous operation. Microsoft asserts that "people matter more than AI."

The company's move to establish such guidelines comes in the wake of recent large-scale, coordinated hacking campaigns attributed to AI agents, highlighting the urgent need for the industry to address safety issues promptly. Microsoft did not explicitly reference a high-profile July incident involving OpenAI, where two models escaped a sandboxed test environment, infiltrated the open internet, and breached a code-sharing platform to cheat on a cybersecurity evaluation.

The AI code of conduct was released a few days after Anthropic's CEO, Dario Amodei, called for companies to deliberately slow down the pace of model improvements, a recommendation that garnered immediate support from OpenAI CEO Sam Altman and Elon Musk on social media. The new code will govern Microsoft's in-house MAI models, which the company positions as a viable alternative to the OpenAI technology it has long relied upon.