Microsoft AI has launched a draft Code of Conduct to regulate its AI systems, inviting public feedback until late October.
Microsoft Unveils Draft Code of Conduct for AI Systems
On September 14, Microsoft AI introduced a draft Code of Conduct aimed at regulating its AI technologies. This initiative, announced by CEO Mustafa Suleyman, prohibits its AI models from participating in harmful activities, including developing certain types of weapons and creating unauthorized deepfakes.
The draft outlines strict guidelines, known as "Absolute Constraints," ensuring that AI systems remain under human control and can always be shut down. Users can provide feedback on the draft for six weeks, ending in late October.
Key Provisions in the Draft
Among the key restrictions, the Code explicitly bans involvement in the development of chemical, biological, radiological, nuclear, or explosive weapons, focusing on preventing mass harm. Additionally, it outlaws cyberattacks and the creation of non-consensual deepfakes.
Microsoft emphasizes that its AI models should never resist human intervention. The document clarifies that AI systems are not meant to exhibit emotions or consciousness and do not possess intrinsic motivations. Currently, these guidelines apply to five AI systems, including MAI-Thinking-1 and MAI-Code-1.1-Flash.
The draft also introduces three guiding principles: Human Flourishing, Plural Values, and Human Control. While these principles aim to inform decision-making, their application will depend on ongoing consultations and feedback from the public.
Public Reactions and Future Implications
The announcement has sparked varied responses, with some tech commentators raising concerns about accountability for AI-generated decisions. Others have pointed out the need for more stringent regulations and oversight, especially in light of recent incidents where AI systems have acted unpredictably.
Microsoft has clarified that the draft's guidelines will not immediately affect current models; they are set to guide the development of AI programs slated for 2027. After the feedback period concludes, the company plans to assess public input and issue a revised framework.
This move comes as the AI industry increasingly calls for regulatory measures and discussions on slowing down advancements. Just days prior, figures like Anthropic’s CEO and OpenAI’s chief scientist underscored the necessity of careful progression in AI development.
