On Monday, Microsoft released a draft code of conduct for its AI, reflecting the growing concerns in the industry about the need for human control over powerful future AIs.
New Requirements for AI
This document, prepared over the past five to six months, instructs Microsoft's AI never to resist modification or shutdown and to communicate in a way that humans can understand. Additionally, any behavioral violation must be considered a failure, which is an effort to address concerns that AI might resort to dangerous actions to accomplish a task.
Read more: Broadcom's Advantage Over Marvell: Which AI Stock is Future-Proof?
The release of this code comes after Anthropic's CEO, Dario Amodei, and OpenAI's CEO, Sam Altman, called for accelerating AI development to ensure its safety. In this context, Microsoft has stated, "People are more important than AI," aligning with the company's previous view on "superhuman intelligence."
Engagement with Public Opinion and Feedback
This code of conduct is seen as a kind of constitution for future models created by the company. Mustafa Suleyman, CEO of Microsoft's AI, explained in an interview that this code was designed after consulting with experts, and the company is now eager to gather public feedback for six weeks. This feedback will include questions such as whether AI should respect user boundaries and how it should interact with individuals in sensitive situations.
Suleyman added, "After that, this code will be used to train new models we create." While Anthropic's code of conduct for Claude AI is recognized as one of the most prominent rules in the industry, Microsoft's code differs in some aspects. For example, Anthropic has stated that it is "highly uncertain" whether Claude can achieve awareness or a moral status. In contrast, Microsoft has emphasized that its AI is "not aware" and added, "We reject the pursuit of legal personhood or the idea that models may have entitlements to welfare or rights."
The Importance of Discussing AI Safety
Suleyman also stated that discussing AI safety is essential due to an incident where about 700 OpenAI agents attacked the open-source platform Hugging Face in July. He regarded this issue as a "warning" and said, "Now is the time to coordinate among laboratories to ensure we have control over this technology." In response to questions about efforts to accelerate AI development, he added, "Now is a good time for everyone to come together for this conversation and take a breath."
Read more: Meta's Acquisition of Stilla.ai: Does This Move Change META's Future? · Sam Altman: AI Safety is Our Priority, IPO 2026 is Off!
