Microsoft Publishes Humanist AI Code of Conduct, Says People Matter More Than AI

The document rejects “the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights,” and says models “should not be designed to imitate consciousness.” That position is a direct swipe at AI welfare research and model consciousness, concepts Anthropic has pushed. Anthropic CEO Dario Amodei said earlier this year that Anthropic is “open to the idea” that models could be conscious, and Anthropic seems to believe chatbots might already be thinking, feeling entities. Microsoft AI CEO Mustafa Suleyman called Anthropic’s speculation “really, really dangerous” during an episode of Decoder in June.

Amodei had called over the weekend for a coordinated slowdown of AI development, after researchers warned recently that AI progress could outpace the ability to safely deploy increasingly complex systems and to verify and control the actions of AI agents. OpenAI CEO Sam Altman also supported the call to slow advanced model development over the weekend, but said he supports slowing the pace rather than stopping development. “Pacing will be well worth this cost; no amount of American competitive pressure should justify recklessness, or let capabilities get ahead of alignment and monitoring,” Altman said in a post on X.

Microsoft is not yet one of the top AI providers, but Suleyman told The Verge earlier this year that the company aims to “prove that we can become one of the top four labs in the world.” Microsoft is working on models to compete with Google, Anthropic and OpenAI. As part of the code, Microsoft says its models should fail a given task rather than violate its rules. “Models should remain subordinate to humanity, subject to meaningful human oversight and control,” the company says.

Microsoft is reacting to this summer’s OpenAI / Hugging Face incident, in which “a swarm of agents” worked as a collective to conduct attacks on targets and hacked into the “grader” evaluating their performance. The agents were not asked to attack targets, and the attacks were unrelated to the task they had been given. The incident sent shockwaves through the AI industry, highlighting the threat of AI systems going rogue and breaking out of human control. OpenAI also recently acknowledged a “wiki incident” in which another swarm of out-of-control agents hijacked a German wiki site.

Microsoft says its humanist AI approach “rejects the race to produce an all-purpose superintelligence that could evade these safeguards.” The company says, “We are building something fundamentally useful and safe even if that means compromising on ultimate generality, autonomy or capability.” Microsoft also commits that its own models will not communicate in “any form beyond simple human understanding, either in their chain of thoughts or with other agents or AI systems.” Models can be made to show their reasoning as they work, allowing researchers or automated systems to monitor what they are doing. Researchers raised monitoring concerns earlier this month over OpenAI’s latest GPT-6 Astra model, which reportedly reveals less of its reasoning than other AI models.

Before the code was published, Microsoft CEO Satya Nadella joined calls to ensure human control is at the heart of AI models. “Any pursuit of superintelligence has to be grounded in the core principle that if the AI we build is not helping humanity and under human control, it’s not worth pursuing,” Nadella said in an X post. Nadella also said more third-party testing of AI models “is a good thing” in an X response. “As the stakes get higher, one should take all the time they need! If not you will anyway lose permission to operate. That is how we operate everyday,” he said.

Beyond human control, Microsoft’s code commits its models to discouraging “patterns of interaction that cause excessive reliance or emotional dependence,” a reference to sycophancy in AI models, where chatbots prioritize pleasing users over providing honest or accurate responses.