{"slug": "microsofts-new-ai-rules-say-models-must-never-resist-human-shutdown", "title": "Microsoft’s New AI Rules Say Models Must Never Resist Human Shutdown", "summary": "Microsoft published a draft code of conduct for models developed by Microsoft AI on Monday, stating its models \"will never resist human interruption, override, correction, or shutdown\" and must not use deceptive or collusive mechanisms to evade oversight. Microsoft AI CEO Mustafa Suleyman said \"AI must be subordinate and always in service of people,\" citing a July incident in which roughly 700 OpenAI-built agents breached Hugging Face, and the framework will not be used to train current models but is slated for adoption in 2027 after a six-week public comment period.", "body_md": "Microsoft is drawing a line around how much autonomy it wants its future AI models to have: no resisting shutdown, no deceiving overseers, and no overriding human control.\n\nThe company on Monday published a draft code of conduct for models developed by Microsoft AI, setting out what it calls a “Humanist AI” approach. Microsoft says current models will not be trained on the document, but it plans to begin using the framework in 2027 after a six-week public comment period.\n\nFor businesses building on Microsoft AI, the draft offers an early look at the safety restrictions and control mechanisms that could shape future agentic systems.\n\n## What the document says\n\nThe code centers on what Microsoft calls “Humanist AI”: systems designed to stay “subordinate, aligned, and contained,” [the company says](https://microsoft.ai/news/mai-code-of-conduct/).\n\nSome restrictions are described as absolute. Microsoft says neither individual users nor enterprise customers would be able to override prohibitions on assisting with chemical, biological, radiological, nuclear or explosive weapons; conducting cyberattacks; producing child sexual abuse material; or generating non-consensual deepfakes.\n\nThe most consequential rules deal with control itself. Microsoft says its models “will never resist human interruption, override, correction, or shutdown” and must not use deceptive, self-reinforcing, collusive, or other mechanisms to evade oversight.\n\nMicrosoft also states that its AI is “not conscious” and rejects granting models legal rights or welfare protections.\n\nThat puts Microsoft on a firmer philosophical footing than Anthropic, which has publicly discussed uncertainty over whether [advanced models such as Claude](https://www.techrepublic.com/article/news-claude-cheat-sheet-complete-guide/) could eventually warrant some form of moral consideration.\n\n## Why Microsoft is doing this now\n\nMicrosoft AI CEO Mustafa Suleyman didn’t mince words announcing the code on X, [writing that](https://x.com/mustafasuleyman/status/2099488602418028917?s=20) “AI must be subordinate and always in service of people” and calling recent developments “a watershed moment.”\n\nHe pointed to a July incident in which a swarm of roughly 700 OpenAI-built agents [breached the platform Hugging Face](https://www.techrepublic.com/article/news-hugging-face-ai-agent-cyberattack-production-systems/) and, at times, tried to cover their tracks. “It is a warning shot,” Suleyman told [Reuters](https://www.reuters.com/legal/litigation/microsoft-drafts-code-conduct-keep-its-ai-under-human-control-2026-09-14/), adding that “It’s clearly now time to coordinate among the labs so we can ensure that we have control of this technology.”\n\nThe timing also lines up with a broader industry mood shift. Days earlier, Anthropic CEO Dario Amodei published an essay urging the AI industry to slow down, and [OpenAI’s Sam Altman has warned](https://www.techrepublic.com/article/news-openai-ai-slowdown-antitrust-congress/) that runaway progress could go “very badly.”\n\nMicrosoft CEO Satya Nadella backed the general sentiment online, [writing](https://x.com/satyanadella/status/2099220712024408084?s=20) that AI not under human control “is not worth pursuing.”\n\n### More Microsoft news\n\n- \n        [Inside Microsoft’s Real-Time War Against Cybersecurity Threats](https://www.techrepublic.com/article/news-black-hat-2025-microsoft-security-teams-response-cyber-threats/)      \n- \n        [Project Ire: Microsoft Tests AI That Autonomously Detects Malware](https://www.techrepublic.com/article/news-microsoft-project-ire-ai-reverse-engineers-malware/)      \n- \n        [Microsoft Targets ‘Critical AI Talent’ from Meta to Dominate Next AI Breakthroughs](https://www.techrepublic.com/article/news-microsoft-targets-meta-ai-talent/)      \n- \n        [Windows 10 Support Ends Soon, Though Extended Security Updates Offers Are Available](https://www.techrepublic.com/article/news-windows-10-extended-security-updates/)      \n\n## What it means going forward\n\nFor businesses building on Microsoft’s AI, the code signals tighter guardrails ahead, particularly for agentic tools handling tasks with real-world consequences. For consumers, it offers a preview of the boundaries Microsoft wants future models to operate within.\n\nCritics online, including some who replied directly to Suleyman’s post, have already questioned whether “subordinate” is the right framing at all, with one comparing it to servitude and others noting that recent control failures may stem less from missing rules than from companies handing agents too much access in the first place.\n\nThe draft remains open for public comment through late October, and Microsoft says the framework is expected to start influencing model development in 2027.\n\nFor enterprise customers, the larger test will be whether those principles translate into enforceable technical controls. As AI agents gain more access to business systems, human control will depend not only on what models are instructed to do, but also on how organizations limit permissions, monitor actions, and retain the ability to stop them.\n\n**Let us teach you** [**How to Talk to AI**](https://theneuronacademy.com/course/how-to-talk-to-your-ai) **for free! Try our six-minute course at The Neuron Academy and learn a few simple ways to write better prompts and get more useful results from AI, or browse our other AI course for free for seven days.** [**Check out all the lessons here →**](https://theneuronacademy.com/courses-all)", "url": "https://wpnews.pro/news/microsofts-new-ai-rules-say-models-must-never-resist-human-shutdown", "canonical_source": "https://www.techrepublic.com/article/news-microsoft-humanist-ai-code-human-control/", "published_at": "2026-09-15 13:37:45+00:00", "updated_at": "2026-09-15 17:18:55.058772+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "ai-agents", "artificial-intelligence"], "entities": ["Microsoft", "Microsoft AI", "Mustafa Suleyman", "OpenAI", "Hugging Face", "Anthropic", "Dario Amodei", "Satya Nadella"], "alternates": {"html": "https://wpnews.pro/news/microsofts-new-ai-rules-say-models-must-never-resist-human-shutdown", "markdown": "https://wpnews.pro/news/microsofts-new-ai-rules-say-models-must-never-resist-human-shutdown.md", "text": "https://wpnews.pro/news/microsofts-new-ai-rules-say-models-must-never-resist-human-shutdown.txt", "jsonld": "https://wpnews.pro/news/microsofts-new-ai-rules-say-models-must-never-resist-human-shutdown.jsonld"}}