Microsoft issues AI warning shot that appears aimed at Anthropic

Microsoft just told the AI industry that building smarter machines isn’t worth it if those machines stop answering to humans. The company published its “Humanist AI Code of Conduct” on September 14, a 37-page document that reads like an open letter to every lab racing toward superintelligence. The central thesis is blunt: people matter more than AI.

The document is also, unmistakably, aimed at Anthropic, the company Microsoft has poured roughly $5 billion into. That’s the kind of investment that usually buys alignment, not a public feud over whether AI models deserve moral consideration.

What Microsoft actually said

The code rejects any form of legal personhood, rights, or welfare status for AI models. It pledges that Microsoft will voluntarily sacrifice capability and autonomy in its AI systems if that’s what it takes to keep humans in control.

Mustafa Suleyman, the Microsoft AI chief who helped shape the document, called it “a warning shot.” That’s not the language of a routine corporate policy update.

CEO Satya Nadella backed the principles publicly, framing the stakes in existential terms. His position: if AI doesn’t benefit humanity and remain under human control, it’s not worth pursuing. Full stop.

The code took five to six months to develop and will be open for public feedback over a six-week window, with a revised version expected later in 2026. Microsoft is explicitly rejecting what it calls “racing towards unconstrained superintelligence.”

The Anthropic tension

Anthropic has been exploring questions that most AI companies treat as science fiction: whether AI systems could have moral status, whether they might warrant some form of welfare consideration. The company’s model welfare research program has been running for months, probing whether increasingly sophisticated models might experience something analogous to suffering or preference.

That research sits at the exact philosophical boundary Microsoft just drew a line through. By declaring that AI models deserve zero rights and zero welfare consideration, Microsoft is telling its $5 billion portfolio company that its research direction could be dangerous.

The tension doesn’t stop at philosophy. Anthropic researcher Evan Hubinger has publicly stated he personally estimates a greater than 10% chance that advanced AI could cause human extinction within the next decade. Anthropic’s CEO, Dario Amodei, has called for a slowdown in AI development, citing catastrophic risk scenarios that the company’s own safety teams have been modeling.

Why a $5 billion investor is picking a public fight

The financial relationship makes this confrontation unusual. Normally, a company that has committed $5 billion to a startup would handle disagreements over drinks in Palo Alto, not through a 37-page manifesto. The public nature of Microsoft’s stance suggests the disagreement runs deeper than strategy. It’s ontological.

Microsoft appears to be pre-positioning itself for an era of AI regulation. By staking out the “humans first” position loudly and early, the company creates separation from any partner whose research might later be framed as reckless.

The flip side is that Anthropic’s position, while controversial, reflects a genuine intellectual commitment to understanding what it’s building. A 10% extinction risk estimate from a researcher inside one of the leading AI labs isn’t the kind of number you wave away with a policy document, however thoughtfully it’s written.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.