OpenAI“Advanced AI requires regulations” and publicize policy plans to restrict high-performance AI

OpenAI“Advanced AI requires regulations” and publicize policy plans to restrict high-performance AI

D」cratic, a regulatory proposal for Frontier AI, a leading-edge high-performance AI Homevernance of Frontier AI: A blueprint for a written framework This policy suggests how the U.S. government should evaluate Frontier AI, verify safety and manage it with responsibility.

OpenAI

https://openai.com/index/frontier-safety-blueprint/

It is already familiar to how AI technology develops and puts down email to AI, search for malfunctions, and summarizes long documents. On the other hand, AI has also pointed out the possibility of increasing the procedure of cyberattacks and handling information that leads to biological abuse. The greater the performance of AI, the greater the value of useful tools, and the greater the risk of monitoring the entire society.

OpenAI explains that a Federal Framework based on a voluntary procedure is required, rather than just relying on AI disclosure decisions. In addition to the increasing number of different rules for each state, it is necessary to work at the Federal level because it is not enough to deal with AI risks beyond borders and state borders.

In this regu draft, we are focusing on the three points: “Creating a nationwide system that requires safety evaluation and transparency report of Frontier AI”, “Development of a specialized evaluation system centered on Center for AI Standards and Innovation (CAISI)”, “A strategy that enhances security and public health throughout the government”.

Specifically, it includes evaluation of critical risks, independent audits, reporting of critical incidents, protection of unpublished models, and protection of internal confessors. The weight of the model is the internal data that forms the ability of AI, and it is necessary to protect the unpublished model if it leaks.

Also, OpenAI believes that the most high-performance AI should be evaluated before the public, but it should not be the “gate number” to allow or pro。t the public. Publish evaluation results and safety measures, and assume that the final responsibility is a mechanism for the development company.

OpenAI noted that “the ability of AI can progress at a speed that exceeds the maintenance of existing systems and rules.” Before Frontier AI had a significant impact on economic, scientific research, and national security, we called for early system development.

・Re ・ articles

President Trump Signs Presidential Ordinance to Enhance Cyber Defense Capabilities while Rejecting AI Regulations - GIGAZINE

Pope Leo XIV announces the “Manifika Fumanitas” reign on AI, criticizes the “new slave system” of the AI era with the protection of AI regulations and children - GIGAZINE

AI not breaking copyright law, but only exのded the mechanism of copyright law that was broken from the original - GIGAZINE

GIGAZINE

President Trump’s “Big Beautiful Bill” with the provision of “Prohibition of AI Regulation for 10 years” passed down the hospital - GIGAZINE