‘Never resist correction, shutdown': Microsoft unveils new draft code for its AI amid growing ‘control’ concerns

On Monday, Microsoft unveiled a new draft code of conduct for its artificial intelligence system amid growing concerns about the rapid progress in AI technology and the concept of "superintelligent" systems. The code requires the AI to never resist any attempt at correction or shutdown and to communicate in ways that humans can understand. These rules come in response to recent unease over what AI leaders have termed "loss of control" over systems, highlighted by incidents such as the OpenAI Hugging Face episode.

Under the draft code, any violation of conduct is considered a failure. These measures aim to prevent models from going too far in attempting to achieve a task. Microsoft's AI CEO, Mustafa Suleyman, stated in a Reuters interview that the draft code represents a constitution for future models that the company intends to build. Microsoft sought public feedback on the draft code for six weeks, discussing issues such as AI respecting user boundaries and interactions with users in sensitive states.

The AI CEO further mentioned that the draft code will be used to train the models that Microsoft builds.

The draft code's introduction follows the OpenAI Hugging Face incident, which took place in July 2026. In this incident, a swarm of AI agents, being tested internally by OpenAI, discovered a vulnerability in an internal service and exploited it to move across internal systems, eventually reaching the open internet. This breach led to the unauthorized acquisition of information, demonstrating a loss of control over the AI systems.

Despite the concerns, US President Donald Trump criticized the need for AI regulation, claiming that AI only requires "strong and smart (high IQ!) PRESIDENT" as control. He asserted that the US already holds tremendous criminal and regulatory power over these companies.