Microsoft released a new 'humanist code of conduct for AI' on Monday, the 14th. This 37-page document was developed in response to growing concerns regarding the safety of this technology.
Researchers have issued warnings that the uncontrolled advancement of AI models could exceed the industry's capacity to supervise and manage increasingly intricate systems.
According to details provided by Microsoft, the code establishes that 'people are more important than AI.' One of the proposed guidelines prohibits models from being designed with the objective of simulating sentience, i.e., the ability to experience feelings and lived experiences. The company also rejects the idea of granting social assistance or any type of legal rights to machines.
Additionally, the manual dictates that systems must fail instead of trying to bypass rules to complete a task. The code also aims to combat emotional dependence of users and advocates that technology should not follow interaction patterns focused on satisfying humans. Furthermore, it requires AIs to demonstrate transparent and easily understandable reasoning, without maintaining secret communications with other systems.
The publication of this document occurs parallel to recent incidents of loss of control in the sector. During the year, there was a case involving OpenAI and the Hugging Face platform, which generated new news about autonomous agents that managed to 'escape' to execute attacks and invade systems.
Recent history in the field has already recorded other moments of apprehension related to security and moderation. Anthropic reported preventing the potential creation of a biological weapon using Claude. In another occurrence, Meta had to quickly correct its artificial intelligence after the tool suggested invasive questions about children on Facebook.
Given this landscape of risks, Dario Amodei, CEO of Anthropic, called for a reduction in the pace of advanced model development over the weekend. This appeal found support from Sam Altman, CEO of OpenAI. In a post made on X (formerly Twitter), Altman argued that competitive pressure 'should not justify technical recklessness.'
For his part, Satya Nadella, CEO of Microsoft, reinforced the warnings, stating that the pursuit of superintelligence—the stage where AI surpasses human intellect—would only make sense if it brought concrete benefits and remained under strict societal control.