Microsoft Unveils AI Code of Conduct Prioritizing Humans Over AI

Microsoft has unveiled a new set of AI development principles that prioritizes human control and safety over expanding artificial intelligence autonomy and capabilities.

Bloomberg reported on September 14 that Microsoft AI researchers published a roughly 15,000-word “Humanist AI Code of Conduct.” The company plans to use it as a kind of constitution for the development of its most advanced AI models.

At the core of the code is the principle that “people matter more than AI.” It states that AI models should not be granted legal personhood or rights, and should not be designed to escape human control or deceive users.

The code requires AI to halt a task entirely if completing it would require violating the rules. The measure is intended to prevent a repeat of security incidents such as an OpenAI model operating without existing safeguards and hacking AI startup Hugging Face.

Microsoft also drew a line in the competition over AI performance. The code says, “Humanist AI rejects the race to build artificial general intelligence that can evade these safeguards,” and says the company will build systems that are fundamentally useful and safe, even if that means giving up some degree of ultimate generality, autonomy or capability.

The framework was prepared by Microsoft AI, the unit led by Chief Executive Officer Mustafa Suleyman. Microsoft removed a clause that had restricted its own development of general AI models when it renegotiated the terms of its contract with OpenAI last year, and is aiming to introduce a cutting-edge AI model as early as next year.

Microsoft said it drafted the code with input from experts in law, ethics and philosophy, as well as corporate officials. The company also plans to gather additional outside feedback. The final principles are set to be applied to the company’s AI model development starting next year.