AI bosses are entering the international political stage one after another! The leader of OpenAI personally went to the United Nations to promote AI standards to promote the commercialization of AI

The Zhitong Finance App learned that OpenAI CEO Sam Altman will officially propose to the global technology industry and politicians to establish globally recognized AI standards when delivering a speech at the UN General Assembly on Wednesday EST, and position himself as an “AI technology middle man” on the question of whether the development of this technology should be slowed down.

Altman, the company's co-founder and CEO, will personally attend an important UN Security Council meeting. Diplomats from various countries will discuss artificial intelligence, application prospects and regulatory guidelines at the conference. Dario Amodei, the most senior director and CEO of Anthropic PBC, OpenAI's strongest competitor, will participate via video link.

Altman, the helm of OpenAI, and Amoday, Nvidia CEO Huang Renxun, Tesla, and SpaceX CEO Musk are frequently moving to the cutting edge of international politics and technology, at a time when cutting-edge AI smart working systems such as Muse and Astra have expanded the scope of complicated and complex commercial tasks that can be automatically completed, which in turn extends computing power requirements from a single question and answer to an extremely complex intelligent workflow that continues to operate efficiently. Global AI computing power demand is expected to usher in a new round of expansion, driving the title of “chip stock trend trend” The Philadelphia Semiconductor Index this week Since then, it has surged nearly 7%, and the increase so far this year is as high as 80%.

AI standards initiative on the UN podium

A representative of OpenAI said at a briefing held before Altman's speech that the company leader will play the role of a “pragmatic artificial intelligence middle man” to push for the formulation of common AI safety standards without hindering innovation or concentrating too much power in the hands of a few companies.

The conference will be held during the annual general debate of the United Nations General Assembly in New York. A major theme this year was concerns about AI getting out of control — concerns were further exacerbated by the resignation of an Anthropic researcher earlier this month to express protest.

UN Secretary General António Guterres spoke on Tuesday warning of the dangers posed by AI and calling for global cooperation. President Donald Trump spoke shortly thereafter, dismissing the idea as a “globalist plot.”

In response to questions about Trump's remarks, an OpenAI spokesperson said the company did not support any specific international regulators. Instead, OpenAI advocates setting basic standards around AI security and implementing controls through democratic mechanisms. They said AI security research institutes in various countries can participate.

An OpenAI spokesperson said that the United Nations can also play a role in discussions on standards, and countries can democratically propose different solutions.

Altman first discussed the idea of establishing a global organization led by the United States in July, and recently explained the idea of establishing such a network in a blog post earlier this week.

Anthropic did not respond to a request for comment on what Amodei would post at the conference.

AI bosses move from the cutting edge of AI technology to the international agenda: global AI application prospects, security assessments, and infrastructure have become business topics

Judging from OpenAI's public statement, one direct reason for participating in international policy discussions is the differences in assessment and reporting systems faced by multinational deployments: different countries may have inconsistent definitions of model capabilities, security incidents, and risk evidence, making comparison and coordination more difficult.

The plan proposed by the leading global AI application company on September 21 covers capability measurement, risk assessment, adequacy of safety measures, and manual review triggers and event notifications in automated AI research. OpenAI clearly defines these elements as a common technical foundation and says that governments decide for themselves whether and how to incorporate them into the legal system. To a certain extent, this also explains Altman's move towards public issues dominated by the international political scene: as models can perform more complex practical tasks, companies need to closely discuss with the government how to evaluate AI capabilities, clarify monitoring procedures, and handle cross-border AI deployment events.

Another issue directly related to business operations is the actual deployment prerequisites for computing power infrastructure facilities and the global AI infrastructure communication framework. On September 16, Nvidia announced the establishment of an AI energy management alliance with Google and Emerald AI to unify technical requirements, performance indicators and operating data sharing methods for flexible electricity facilities, and explore faster grid access paths; the company clearly stated that electricity has become an important constraint on the expansion of AI infrastructure in the US.

Huang Renxun also discussed AI safety responsibilities and regulatory cooperation at the G20 Innovation Ministers Meeting on September 2. This is a different event from this UN conference. For stock market investors focusing on the topic of AI computing power, these public topics connect the acceleration of the commercialization process of AI models/cutting-edge AI agents, data center operation time and capital expenditure: the performance of core AI chips and data center CPUs and memory chip components determines computing power, power access and project delivery determine usable capacity, and specific evaluation and deployment requirements enter the enterprise's operating process and cost structure.

Frontier AI systems such as Muse and Astra expand the range of commercial tasks that can be automatically completed, and also extend computing power requirements from a single question and answer to an intelligent workflow that operates continuously. Meta also revealed that Muse runs on a dedicated cloud virtual machine equipped with a browser, and can continue to process tasks even if the user closes the application. Improved model efficiency and improved task success rate can allow more jobs that were originally too expensive to enter the AI system. Therefore, the impact of the popularity of smart devices on the industry includes not only the demand for AI chips, but also the expansion of supporting CPUs, storage, interconnection, and power supply systems.

Looking at the underlying architecture, a complex task may include multiple rounds of inference, tool call, code execution, and result verification: GPUs are responsible for model matrix computation, long context pre-filling increases input processing load, and continuous decoding and concurrent sessions increase requirements for computational throughput, memory bandwidth, and key-value cache capacity; the CPU undertakes tool execution, sandbox operation, and task orchestration. The expansion of the number of users, task coverage, and parallel workflows forms the basis for the growing demand for GPUs, CPUs, storage, and networks. This is also an important logic for the market to popularize smart device applications and turn it into growth expectations for computing power vendors such as Nvidia and AMD.