Microsoft Unveils Draft Code of Conduct for AI Projects

Microsoft has released a draft 'Code of Conduct' for its artificial intelligence (AI) projects, aimed at keeping powerful AI models under human control. The code is part of a growing concern across industries that companies must contain and align such powerful forces.

The document requires Microsoft's AI models to never resist correction or shutdown, communicate in ways that humans can understand, and consider any conduct violation as a failure. It also specifies that the models will not obscure their underlying nature as an AI, nor claim interiority, feelings, experiences, or a soul.

Furthermore, the code prohibits AI models from being deceptive, impersonating humans, moderators, or other forms of authority in forums, defying boundaries like emotional or social, and using unauthorized platforms for persistent memory or context. The new rules aim to prevent AI-enabled cyberattacks and prioritize AI safety in model development.

The announcement comes as industry leaders, including Dario Amodei and Sam Altman, have renewed calls for pacing AI's development to ensure its safety. Microsoft's draft code represents a system for future models made by the company and was created with help from experts in AI, law, ethics, philosophy, linguistics, and public policy.