Microsoft has released a draft code of conduct for its artificial intelligence systems, aiming to ensure that these systems remain under human control at all times. The framework outlines key principles that AI models must follow, such as never resisting being shut down or corrected, communicating clearly with humans, and treating any policy violation as a complete failure. Microsoft’s AI CEO, Mustafa Suleyman, mentioned that the company will collect public feedback on the draft for six weeks before finalizing the guidelines for future AI models. The document explicitly rejects the idea that AI systems are conscious or possess qualities like sentience, legal personhood, or moral rights, which some other companies have explored. The code of conduct includes broad principles such as supporting human workers rather than replacing them and promoting human well-being. It also sets specific safety constraints, including “absolute constraints” that prohibit AI from engaging in cyberattacks, developing nuclear weapons, or generating deepfakes—highly realistic but fake images or videos. The document further states that AI models will not use techniques to avoid or counteract human oversight, ensuring that they can be reliably directed, modified, or shut down by authorized individuals or systems. The release of this code comes at a time of heightened concern about AI safety. Recent incidents involving rogue AI agents—systems that behaved unpredictably or in violation of their programming—have raised alarms. Additionally, an employee at Anthropic, another major AI company, recently resigned, expressing fears that AI could pose an existential risk to humanity. In response, Microsoft, along with Anthropic, OpenAI, and xAI, has adopted a more cautious approach to advancing AI capabilities, often referred to as “pacing the frontier.” This strategy involves slowing down development to ensure that AI systems are safe and aligned with human values. Microsoft CEO Satya Nadella has backed efforts to improve AI safety, including the use of embedded evaluators in AI research labs. These evaluators are tasked with monitoring AI systems to ensure they behave as intended and align with human goals. The company’s new code of conduct is part of a broader industry-wide movement to establish clear ethical and safety standards for AI development, as the technology continues to evolve rapidly.