Microsoft AI Code Project: What the Available Materials Confirm

The available materials do not confirm that Microsoft presented an AI code project. They do not establish the content, status, or discussion timeline of such a document.

Status of the Microsoft AI Code Project

The available materials contain no confirmation that Microsoft presented this project, planned to discuss it with stakeholders, or release an updated version later in 2026.

Without a confirmed text, it is impossible to establish what provisions the document contains or which products or functions they concern. It is also impossible to assess whether it would guarantee specific behavior by Microsoft products.

Humanist AI and the Role of Humans in Decision-Making

The available materials do not confirm that the Microsoft project defines Humanist AI as an approach in which AI acts in people’s interests and remains under their control.

They also do not confirm that the document establishes rules for models that use tools or act as agents, including limits on their authority, independently changing access, or choosing goals.

Information about the ability to stop a system, correct its behavior, or restrict its functions in connection with this project has not been confirmed.

No Confirmation of Microsoft’s Position on the Status of AI Models

There is no confirmation that the project opposes legal personhood or rights of their own for AI models, or that it contains rules on how models should describe consciousness, feelings, desires, and motives.

The available materials do not support attributing to the project any conclusions about how models should describe experiences or persuade users.

Nor has the claim that the code prohibits encouraging excessive emotional attachment to AI been confirmed. An emotionally charged response does not in itself prove that a model has feelings.

AI Shutdown, Correction, and Authority

The materials do not confirm that the project prohibits models from resisting intervention, correction, or shutdown, independently expanding their authority, or setting themselves unspecified goals.

There is also no confirmation that the document requires a model to abandon a task following a serious violation or defines what constitutes such a violation.

The available materials contain no information about shutdown procedures, retaining human control, or rules governing communication between AI agents.

Dangerous Capabilities and Limits on Autonomy

There is no confirmation that the project establishes special restrictions for models in the areas of weapons of mass destruction, child safety, or large-scale malicious manipulation.

Nor has the claim that Microsoft is prepared to limit the autonomy, versatility, or capabilities of models for the sake of safety been confirmed.

The available materials do not confirm the statement that Microsoft does not intend to participate in a race for superintelligence at any cost. No conclusions about when superintelligence might emerge or about the intentions of other developers can be drawn from it.

Public Discussion and the Next Version

The six-week duration of the public discussion and the anticipated release of an updated edition later in 2026 have not been confirmed by the available materials.

There is no confirmed information that the document was opened for discussion or that Microsoft planned to revise it.

What the Project Means for Companies and Specialists

Because the project’s content has not been confirmed, the provisions presented in the article cannot be considered Microsoft recommendations. Companies should assess the specific product and the requirements applicable to it.

When evaluating an agentic system, it is worth determining what actions it performs, who assigns it authority, and how the user can stop it from operating. These questions can be considered independently of the Microsoft project.

Brief Answers About the Microsoft AI Code Project

Is the Microsoft AI code already in effect?

The available materials do not establish whether such a document exists or what its status is.

What main principle does Microsoft propose?

The materials do not confirm what Microsoft specifically established in the project.

What should a model do when faced with a dangerous request?

The available materials do not make it possible to establish the refusal rule provided for in the document.

When was an updated version expected?

A date later in 2026 has not been confirmed by the available materials.