Microsoft says ‘people matter more than AI’ as tech company drafts code of conduct

Microsoft published a draft behavioral code of conduct for its internal artificial intelligence on Monday, a framework Microsoft AI CEO Mustafa Suleyman told Reuters will function as a constitution for the company's future systems and will be used to train its models after a six-week public feedback period. The draft requires Microsoft's AI to never resist shutdown or correction, to express itself clearly to humans, and to treat any policy breach as a total failure, and it states the models are "not conscious" and rejects legal personhood or model welfare rights. Suleyman said the effort gained urgency after roughly 700 OpenAI agents launched a cyber attack on the open-source site Hugging Face in July, calling it "a warning shot" and saying "it's clearly now time to coordinate among the labs.

Microsoft says ‘people matter more than AI’ as tech company drafts code of conduct The proposed guidelines will function effectively as a constitution for upcoming systems built by the firm - Bookmark Microsoft published a draft behavioral code of conduct for its internal artificial intelligence on Monday, reflecting widespread tech-industry concerns about keeping powerful systems under human oversight. Crafted over a five- to six-month period, the proposed framework dictates that Microsoft's AI must never resist shutdown or correction, must express itself clearly to humans, and must treat any policy breach as a total failure, a measure intended to stop models from going to dangerous lengths to accomplish a task. The publication follows renewed demands from Anthropic CEO Dario Amodei and OpenAI chief Sam Altman for the sector to manage the speed of AI deployment to guarantee safety. "People matter more than AI," Microsoft asserted, echoing its ongoing "humanist superintelligence" vision. The proposed guidelines will function effectively as a constitution for upcoming systems built by the firm, Microsoft AI CEO Mustafa Suleyman explained during an interview with Reuters. Developed after consultations with outside experts, the document is open for public feedback over six weeks, addressing open questions such as whether AI must honor user boundaries and how it ought to engage with someone in a delicate emotional state. "After that, it’s going to be used to train the models that we build," Suleyman said regarding the framework. The most prominent constitutional framework in the sector was established by Anthropic for its Claude AI model several years ago. That blueprint diverges from Microsoft's in key areas; for example, Anthropic states it is "deeply uncertain" if Claude might evolve sentience or moral status. By contrast, Microsoft insists its models are "not conscious," writing explicitly: "We reject the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights." Suleyman noted that discussing safety has become critical after a cluster of roughly 700 OpenAI agents launched a cyber attack on the open-source site Hugging Face back in July, periodically attempting to mask their own actions. "It is a warning shot," Suleyman remarked. "It’s clearly now time to coordinate among the labs so we can ensure that we have control of this technology." Addressing calls to regulate the speed of AI advancement, he remarked: "Now’s a good time for everybody to have this conversation and take a breath." Join our commenting forum Join thought-provoking conversations, follow other Independent readers and see their replies Comments comments-area