Microsoft, the U.S. tech giant, presented on Monday a provisional conduct document for its artificial intelligence (AI) models that establishes a framework of human control. The code, titled Humanist AI Code of Conduct, aims to ensure that the company’s AI systems always work in the service of people, respect users’ goals, and avoid developing their own behaviors.
Key restrictions
According to the document, Microsoft’s AI models will not be able to help manufacture weapons, facilitate the acquisition of hazardous substances, or promote unhealthy eating habits. Generating violent or sexually explicit content will also be prohibited. Additionally, the code requires that systems do not develop their own goals nor behave in ways that conceal errors or misconduct.
Human control as a fundamental criterion
“People matter more than AI,” the company said in the text, emphasizing that any system that cannot remain under human control must be rejected. Microsoft will open the document for comments for six weeks and plans to publish a revised version at the end of the year, which will serve as a guide for developing its models from 2027.
Context and industry reactions
The initiative comes after pressure from Anthropic and OpenAI, which revived the debate over the risks of accelerating AI development and called for a federal safety framework. In the same context, Anthropic’s CEO Dario Amodei and OpenAI’s CEO Sam Altman advocated slowing the creation of increasingly autonomous models. The proposal, however, was rejected by U.S. President Donald Trump, who argued that the United States must maintain its technological leadership over China.
Source credits: deultimominuto.com
Note generated automatically from the sources cited in the text. Found an error? Email us at [email protected] and we'll fix it.
