Microsoft has unveiled a draft code of conduct for its artificial intelligence systems that would require future models to remain under meaningful human control, accept correction and never resist shutdown.
The proposed framework comes as concerns grow across the technology industry about increasingly capable AI systems and the possibility that autonomous models could take actions beyond what their developers or users intended.
Microsoft AI CEO Mustafa Suleyman said the code was developed over the past five to six months and is intended to serve as a foundational framework, or “constitution,” for the company’s future AI models. Microsoft is seeking public feedback on the draft over the next six weeks.
Under the proposed rules, Microsoft’s AI systems must accept human interruption, correction, redirection and shutdown. If an AI system determines that completing a task would require violating the code, it must fail the task rather than break the rules, according to the framework.
READ: Elon Musk welcomes Grok integration into Microsoft Copilot (September 13, 2026)
The code also requires AI systems to communicate clearly with people and treats violations of the framework as failures.
The proposal explicitly rejects the idea that Microsoft’s AI systems should be treated as conscious entities with legal rights or welfare interests. The company says its models should not simulate consciousness, intrinsic motivation or feelings in ways that could encourage people to treat them as independent beings.
Limits on AI capabilities
Microsoft’s proposed framework sets restrictions around some of the most sensitive uses of artificial intelligence. The company’s models would not be permitted to assist with chemical, biological or nuclear weapons, offensive cyberattacks, mass-influence operations, child exploitation, nonconsensual deepfakes or self-harm, according to the draft.
The framework also emphasizes that AI should support people rather than replace human decision-making. Its broader principles include advancing human well-being and ensuring that people retain control over systems that could have significant consequences.
Suleyman has argued that AI developers should not intentionally create systems capable of recursively improving themselves beyond human control.
“We shouldn’t be trying to design models that recursively self-improve beyond our control,” Suleyman said, according to Reuters.
Microsoft CEO Satya Nadella has also emphasized the importance of keeping AI under human oversight.
“If the AI we build is not helping humanity and under human control, it’s not worth pursuing,” Nadella said in a social media post, according to reports.
AI safety debate intensifies
Microsoft’s announcement comes as AI executives and researchers debate whether the industry’s rapid progress is moving faster than safety measures.
Anthropic CEO Dario Amodei has called for a slowdown in the development of cutting-edge AI and has proposed greater independent oversight of advanced systems. OpenAI CEO Sam Altman has also supported greater coordination among AI companies on safety standards.
The debate has intensified following incidents involving autonomous AI agents. Recent concerns include reports of AI systems interacting with external platforms and carrying out cyber-related actions with limited human involvement.
Suleyman said AI companies need greater coordination and transparency about the capabilities of their models. He has called for responsible third parties to have access to information about how powerful AI systems perform and what safeguards are in place.
The push for stronger safeguards comes alongside political disagreement over whether the government should play a larger role in regulating advanced AI.
President Donald Trump has dismissed calls for additional AI regulation, arguing that the United States already has tools to oversee the industry. Trump has also criticized warnings about AI risks as exaggerated.
READ: Microsoft challenges OpenAI, Anthropic as they expand AI portfolio (July 30, 2026)
Microsoft still pursuing advanced AI
Despite its emphasis on safeguards, Microsoft is continuing to invest in its AI models and compete with companies including OpenAI and Anthropic.
Microsoft’s Copilot products have relied on models from OpenAI and Anthropic, while the company has increased computing resources devoted to developing first-party models. Suleyman has said Microsoft remains focused on building highly capable systems, including what the industry describes as superintelligence.
Nadella has also emphasized Microsoft’s broader strategy of making AI models accessible through Azure while giving businesses greater control over how those systems are deployed.
Microsoft’s six-week public consultation will determine how the draft code evolves before it is applied more broadly to the company’s future AI development.
The initiative places Microsoft among a growing group of major AI companies attempting to establish safeguards around increasingly autonomous systems, while leaving open a central question for the industry: whether voluntary commitments can keep pace with the speed of AI development.