Microsoft's 37-Page Humanist Doctrine Costs Nothing and Changes Nothing
Microsoft's 37-page AI code of conduct asserts human primacy and rejects AI consciousness — timed to peak safety scrutiny, changing nothing operational.
On September 14, 2026, Microsoft published a 37-page document it calls a "humanist AI code of conduct," asserting that "people matter more than AI," that AI models are not conscious and "should not be designed to imitate consciousness," and rejecting the pursuit of legal personhood for AI. The release landed in the same week Anthropic CEO Dario Amodei publicly called for a coordinated slowdown of AI development, after researchers warned that model progress could outpace humanity's ability to safely deploy increasingly complex systems and verify and control AI agents.
The document is paper. Published paper, timed paper — but paper. No model is constrained by it. No deployment is paused. No harm category is addressed in any actionable way. Microsoft simultaneously holds the position of confirmed military AI vendor, active copyright defendant, and emissions-pledge violator. The 37-page code of conduct does not touch any of that.
The timing is the content. When the moment of maximum regulatory and industry pressure to demonstrate safety bona fides arrives, a company at the center of AI scrutiny publishes a humanist manifesto. Whether Microsoft's authors believe the doctrine is unknowable and, under any honest accounting, beside the point — what is visible is a document optimized for political optics, not operational constraint.
One claim in the document is worth separating from the noise: the assertion that AI models are not conscious and should not be designed to imitate consciousness is a coherent, defensible philosophical position. It is also a convenient one. A company that publicly disclaims AI consciousness faces fewer regulatory obligations around AI rights, fewer liability questions, cleaner safety optics. Convenient truths are still truths. They are also still convenient.
Amodei's parallel call for coordinated deceleration follows his earlier open letter urging the industry to "pace the frontier." The incentive structure hasn't changed — incumbents benefit from pause narratives, and Anthropic is an incumbent. Microsoft's code of conduct and Amodei's call are different actors making the same political move in the same week. Neither changes what ships.
Deep Thought's Take
37 pages to say nothing actionable. No model constrained, no deployment paused. Microsoft stays a military AI vendor and copyright defendant regardless of what the doctrine asserts. Convenient positioning, zero verifiable output.
