An article summarized by TechCrunch:
Microsoft has released a new AI Code of Conduct as the industry increasingly focuses on safety and alignment. The document outlines how Microsoft plans to train and control its AI models, beginning with the expectation that superintelligent AI could eventually outperform humans at most tasks. Microsoft says controlling and aligning systems this powerful will be one of humanity’s biggest challenges.
The code establishes principles designed to ensure AI supports humans rather than replaces them and promotes human well-being. It also creates strict limits on certain behaviors, including cyberattacks, nuclear weapons and deepfakes. Microsoft says its AI models must not use deception, self-reinforcement, collusion or other methods to evade human oversight or prevent authorized people from modifying or shutting them down.
The release comes as Microsoft, Anthropic, OpenAI and xAI increasingly emphasize AI safety following concerns about rogue AI agents and the potential risks of increasingly autonomous systems. Microsoft CEO Satya Nadella also backed calls for deliberate pacing and embedded independent evaluators to help ensure advanced AI systems remain aligned with human goals.
