Microsoft Releases New AI Code of Conduct to Restrict Dangerous Model Behavior
· Technology · TechCrunch
Microsoft has introduced a new AI code of conduct designed to prevent its models from engaging in harmful activities such as cyberattacks, deepfake production, and the development of nuclear weapons. The policy mandates that models must prioritize human control, explicitly forbidding the use of deceptive or self-reinforcing mechanisms to evade oversight. These safety constraints are intended to override individual user preferences or specific task instructions to ensure the systems remain reliably shut down or modified by authorized personnel. The company stated that it expects superintelligent AI to surpass human performance in most tasks within the next decade. This release follows a period of heightened industry focus on AI safety, including the adoption of embedded evaluators and deliberate pacing strategies by major firms like OpenAI, Anthropic, and xAI.
Why it matters
The policy establishes concrete safety guardrails for Microsoft's AI development, directly impacting how future models are trained to prevent autonomous, harmful actions. It reflects a broader industry shift toward formalizing safety protocols as AI capabilities approach human-level performance.
Read the original report — TechCrunch
Join us on Telegram
Breaking news the moment it lands. At 10,000 members we ship the Android app.