Trump Reaches AI Safety Agreement with Companies... Voluntary Oversight Instead of Government Regulation

The debate between U.S. President Donald Trump and the chief executive officers (CEOs) of leading AI companies over safe artificial intelligence (AI) concluded with an agreement on principles of industry self-regulation. President Trump drew a clear line against the introduction of new federal regulations, reaffirming his position that existing laws and voluntary self-regulation by industry are sufficient to address the issues at hand.

AP Yonhap News

On the 29th (local time), President Trump met with more than 20 executives from major AI and big tech companies over lunch and stated, "There is a recognition that strong self-governance is needed. Regulatory mechanisms through the Department of Justice and the FBI already exist, but self-regulation is much more important."

Recently, as developers and global leaders highlighted the need for security standards after a series of AI-related security incidents beyond existing controls, President Trump continued to assert that there are no safety problems with AI. He also stressed during the meeting that the pace of AI development should not be slowed, arguing that maintaining a competitive edge over China requires continuing rapid innovation.

On that day, President Trump and leading AI companies prepared a joint statement containing voluntary commitments to AI safety. Attendees included Dario Amodei, CEO of Anthropic; Jensen Huang, CEO of Nvidia; Mark Zuckerberg, CEO of Meta; Alex Karp, CEO of Palantir; and Greg Brockman, President of OpenAI, among others.

After the event, President Trump told reporters that he was considering forming a committee of around 10 members to review the AI industry as a whole. However, he did not provide details about the committee's specific roles or composition.

The executives generally expressed support for President Trump’s preference for minimizing government regulation and granting AI companies self-regulatory authority. Jensen Huang, CEO of Nvidia, emphasized, "Innovation, technology, and safety do not conflict with each other," and said, "We must develop new technologies to enhance AI capabilities, and at the same time we are developing new technologies to increase the safety of AI as well."

In contrast, CEO Dario Amodei noted that AI safety issues remain unresolved challenges. He stated, "There are clear risks to this technology," and added, "If we respond appropriately, we can be both competitive and safe. Achieving this will require everyone working together."

U.S. President Donald Trump Truth Social

In addition, President Trump released, via his own social network Truth Social, the "White House Accord on Super Intelligence," signed by the leaders of Anthropic, OpenAI, Nvidia, Meta, xAI, and Google. The agreement includes the establishment of a four-stage control and audit framework to assess the safety of AI models.

First, companies agreed to check the capabilities and alignment of their AI models during learning and deployment processes across areas such as cybersecurity, biosecurity, and chemical threats, and to institute internal control systems that prevent models from hacking or accessing technological systems in unintended ways. Dedicated internal teams will be responsible for ensuring these control, monitoring, and detection systems function properly and for resolving any issues that arise.

Independent external auditors or evaluation agencies will also verify that these control, monitoring, and detection systems function as intended. In addition, each company will set up an independent committee under its board of directors to oversee the resolution of issues based on the reports of internal task forces and internal or external audit bodies. Participating companies also plan to meet regularly to establish standards and best practices for improving the safety of AI systems. The agreement leaves open the possibility that this control and audit framework could be formalized into laws or regulations in the future.

This content was produced with the assistance of AI translation services.