Microsoft has released a 37-page humanist AI code of conduct to address safety concerns regarding the speed of model development. This move follows comments from Anthropic CEO Dario Amodei, who recently advocated for a coordinated slowdown in the industry after researchers warned that progress might outpace safe deployment and control capabilities. The document explicitly states that people matter more than artificial intelligence and clarifies that models lack consciousness. Consequently, Microsoft rejects the idea of designing systems to imitate human awareness or pursuing legal personhood for these technologies.
The policy serves as a direct response to growing unease about the risks associated with increasingly complex autonomous agents. By setting these boundaries, the company aims to prevent the development of systems that could operate beyond human oversight or verification. This stance prioritises human oversight over the unchecked expansion of model capabilities.
- The code rejects legal personhood for AI systems.
- Models must not be designed to imitate consciousness.
- Development speed is subject to safety verification limits.
