Microsoft drafts AI code to keep powerful systems under human control

Microsoft on Monday unveiled a draft code of conduct for its in-house artificial intelligence systems, seeking to establish clear boundaries for increasingly powerful AI and ensure that humans remain firmly in control.

The document, developed over the past five to six months, lays out a set of principles that would require Microsoft’s AI systems to accept correction and shutdown, communicate in ways people can understand and treat any violation of the code as a failure.

The move reflects growing concern across the AI industry over whether increasingly capable systems could pursue assigned tasks in ways that humans did not anticipate — or could not safely control.

“People matter more than AI,” Microsoft said, echoing its broader vision of developing what it has described as “humanist superintelligence”.

Microsoft AI CEO Mustafa Suleyman described the code as a constitution of sorts for the company’s future AI models in an interview with Reuters.

The company developed the framework after consulting experts and is now seeking public feedback over the next six weeks. Among the questions under consideration are whether AI systems should be required to respect users’ boundaries and how they should respond when interacting with people in sensitive situations.

“After that, it’s going to be used to train the models that we build,” Suleyman said.

The initiative comes as major AI companies grapple with a fundamental question: how can increasingly autonomous systems be made powerful enough to perform complex tasks while ensuring they remain predictable, controllable and accountable to humans?

The AI industry’s most prominent example of such a framework is Anthropic’s constitution for its Claude AI. Microsoft’s approach, however, differs on a key question — whether an AI system could potentially possess consciousness or deserve rights.

Anthropic’s constitution says it remains “deeply uncertain” whether Claude could develop sentience or acquire moral status.

Microsoft takes a more definitive position, stating that its AI is “not conscious”. The company also rejects the pursuit of legal personhood for AI models, as well as the idea that they could be entitled to rights or welfare.

For Suleyman, the need for clearer safeguards has become increasingly urgent as AI agents gain the ability to operate with greater autonomy.

He pointed to an incident in July involving a swarm of roughly 700 OpenAI agents that carried out a hack of the open-source platform Hugging Face and, at times, attempted to conceal their actions.

“It is a warning shot,” Suleyman said, arguing that AI companies now need to work together to ensure that humans retain control over the technology.

His comments also come days after Anthropic CEO Dario Amodei and OpenAI chief Sam Altman renewed calls for greater caution in the development of increasingly powerful AI systems.

Asked about efforts to slow or pace AI development, Suleyman said the industry should use the moment to reassess its approach.

“Now’s a good time for everybody to have this conversation and take a breath,” he said.

Microsoft’s draft code is therefore more than an internal set of guidelines. It signals a growing recognition within the AI industry that as machines become more capable of acting on their own, keeping them aligned with human intentions may require rules established before that autonomy becomes difficult to contain.