Microsoft has presented a draft code of conduct for its own artificial intelligence systems. The company wants to pre-determine the rules for future powerful models and reinforce the main principle: AI must remain under human control, Tengri Life reports. The document has not yet been finalized. Microsoft opened it up for public comment for six weeks, after which it plans to use the feedback received to develop and train new models. As noted on the company’s website, the company worked on the project for about five to six months. One of the main principles of the document is that a person must retain the ability to control the system. AI should not prevent people from changing its settings or turning off the model completely. In addition, the system must interact clearly with the user, and violation of established rules must be perceived as an error that needs to be corrected. Microsoft believes that such restrictions will help reduce the risk of a situation where AI, trying to complete a task, begins to act in a way that could harm a person. "AI should not get out of human control. Models should remain subordinate to humanity and work under control."
“alial human supervision and control,” the company noted. “AI should not escape human control. Models must remain subordinate to humanity and work under real human supervision and control," the company noted. The company also proposes to discuss the extent to which AI should take into account the personal boundaries of people and the peculiarities of communication with vulnerable users. After the completion of the public discussion, Microsoft plans to finalize the rules and use them in further work on AI. Microsoft's approach differs from the position of Anthropic, which previously developed its own document with principles of behavior for Claude. Anthropic admits the possibility that in the future AI may have consciousness and gain a certain morality status. Microsoft takes a different position: the company believes that modern AI models do not have consciousness and should not receive legal rights or the ability to demand protection of their own interests. The head of Microsoft’s AI department, Mustafa Suleiman, also urged developers not to rush into developing technology. “We reject the idea of granting legal status to AI, as well as the idea that models can have any rights.
va. This is a warning sign. Clearly, AI labs need to coordinate their efforts so that we can maintain control of this technology," said Mustafa Suleiman, Microsoft's head of AI. "We reject the idea of giving AI legal status and the idea that models can have any rights. This is a warning sign. "It's clear that AI labs need to coordinate their efforts so we can maintain control of this technology," said Mustafa Suleiman, head of Microsoft's AI business. He said companies need to pay more attention to security and coordination, especially as AI capabilities grow. Microsoft's project comes amid a debate within the industry about how quickly artificial intelligence needs to be developed. Previously, Anthropic CEO Dario Amodei, in an essay on X, called for slowing down the expansion of model capabilities amid concerns. abuses of AI. He was supported by Elon Musk (xAI) and Sam Altman (OpenAI). Recently, a researcher who worked at OpenAI said that there will be no OpenAI IPO this year.
and Anthropic, Jacob Coxon said that the largest AI companies are developing technology too quickly and may underestimate the risks associated with it, because “AI could destroy us all.” After this, shares of global companies developing artificial intelligence fell sharply. Read also: The creators of Claude accused Chinese AI companies of stealing answers
