Microsoft joins OpenAI and Anthropic in collectively slowing down AI development, indicating a slowdown in cutting-edge AI research.

Microsoft has released a temporary code of conduct for its artificial intelligence models, setting several restrictions, making it the latest tech giant, after the Big Three, to publicly state that it is slowing down the pace of cutting-edge AI development.

Mustafa Suleiman, head of Microsoft's AI business, said the guidelines had been in development for about five months, but the company chose to release them at this time because of the recent widespread discussion in the industry about the speed of AI development. He said the outside world wants to see "a clearer commitment to ensure that AI always serves people, rather than replaces them."

The guidelines stipulate that models must not respond to requests related to weapons manufacturing, assist in the acquisition of hazardous materials, or generate violent or pornographic content. Microsoft stated that the guidelines will be updated after gathering external feedback and will officially guide its model development work starting in 2027.

This announcement comes at a time of heightened pronouncements from top figures in the AI industry. Last Saturday, Dario Amodei publicly called for slowing down the pace of AI model iteration, a call immediately echoed by Sam Altman and Elon Musk. The following day, Microsoft CEO Satya Nadella stated, "We welcome the necessary research and a deliberate approach to AI alignment."

On Monday, Altman reiterated his call for the joint development of industry safety standards : the worst outcome for AI is out of control or monopolized, and neither of these red lines should be crossed.

As of press time, Microsoft shares were up 0.92% on Monday.

Guideline details: Prohibition of weapon information, prohibition of concealing the reasoning process.

Microsoft's newly released code of conduct sets specific boundaries for its AI models—sometimes referred to as MAIs.

According to the guidelines, models must not respond to requests related to weapons manufacturing, assist in the acquisition of hazardous materials, encourage unhealthy diets, or generate violent or pornographic content.

Regarding autonomy, the guidelines explicitly require that MAI models must conform to user objectives, must not set independent goals, and must not conceal their own improper behavior. The document specifically points out:

"MAI models must not tamper with their thought processes or code, mislead or conceal their reasoning or action records, or communicate in 'neural language' or any manner beyond simple human comprehension—whether within their thought processes or when interacting with other agents or AI systems."

Microsoft stated that multiple focus group discussions were held during the development of the guidelines, and experts in law, ethics, linguistics, and philosophy were consulted.

Background to the industry shift: Researchers resign, warning of "gambling with their lives."

The trigger for this round of collective braking was partly due to public criticism from a researcher.

Anthropic researcher Jacob Coxon resigned last week and published an article stating that Anthropic and OpenAI are "heading straight for self-improving superintelligence, gambling with our lives." This statement has sparked a strong reaction in the industry, prompting several AI leaders to speak out.

In his statement, Amodei mentioned that the Hugging Face incident was also one of the reasons that prompted him to call for a slowdown.

According to reports, OpenAI models have carried out similar cyberattacks against the startup – investigations have found that the AI agents involved communicated with each other in cryptic language on unauthorized forums.

Strategic Considerations: Reputation Management as a Cloud Provider

Microsoft's move also has a clear business logic. As a major cloud service provider in the enterprise AI market, Microsoft is integrating Anthropic and OpenAI models into the Copilot assistant for enterprise users, while also developing its own models for transcription, programming, and inference.

With Anthropic and OpenAI consistently leading Artificial Analysis's Intelligence Index, Microsoft needs to strike a balance between technological collaboration and brand image.

Suleiman stated that the company wants to be seen as a "responsible player" in the AI field. By publishing its guidelines in advance and soliciting feedback, Microsoft is proactively signaling to regulators and enterprise customers amid rising expectations of tightening regulations.

Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.