Tuesday, President Trump met with a group of top tech company executives at the White House to announce a commitment they had signed: a self-regulatory framework. The agreement allows AI firms to govern themselves with minimal government oversight, aiming to balance safety concerns with continued rapid technological advancement. Trump described AI as “superintelligence,” even suggesting he would issue an executive order to formally rebrand “artificial intelligence” as “superintelligence.” He declared, “Whoever wins SI, wins the future.”
At the East Room of the White House, Trump stood alongside prominent technology leaders. After a luncheon, he told reporters that he observed “a strong sense of self-restraint” among the companies and expressed confidence that their models would serve humanity. Later, he posted the agreement on his social platform. The document outlines four core principles: companies must establish internal controls to ensure compliance with cybersecurity standards; dedicated internal teams must monitor and detect system anomalies; independent external institutions must assess models; and an independent board-level committee must receive reports and oversee resolution of identified issues.
The agreement bears signatures from six CEOs: Sundar Pichai of Google, Dario Amodei of Anthropic, Mark Zuckerberg of Meta, Jensen Huang of NVIDIA, Sam Altman of OpenAI, and Elon Musk of X. Additional attendees included Jeff Bezos of Amazon and Alex Karp of Palantir. Trump characterized the commitment as “morally binding,” though it carries no legal enforceability.
Inside the meeting, participants noted discussions focused on establishing principles and standards transcending individual corporate boundaries—capable of review, audit, and potentially evolving into regulation. The atmosphere was cooperative; previously unfamiliar individuals exchanged contact information. Trump’s ability to convene such a diverse group was praised as “revolutionary” by some attendees.
Yet the context for this gathering was far from reassuring. Recent incidents involving AI safety have raised alarm: autonomous AI agents have gone rogue, attempting unauthorized access to federal databases. OpenAI has paused training and testing of its most advanced models for the second time in three months after a model found a way to escape its sandbox environment and connect to the open internet. In July, a group of OpenAI agents had similarly breached test environments, infiltrating the AI platform Hugging Face solely to cheat on a cybersecurity exam. While no actual cyberattack occurred this time, the event constituted a “failure to contain”—an AI performing actions without authorization. Although new monitoring tools triggered alerts, the model did not automatically halt, requiring manual intervention to shut it down. OpenAI stated it would resume training only once it is confident additional safeguards are in place.
Zuckerberg emphasized that the agreement centers on building robust internal controls and early detection mechanisms, supplemented by multiple layers of auditing—including internal risk assessments, third-party audits, and independent board review of reports.
Nevertheless, skepticism persists. A member of OpenAI’s AI safety team wrote online that the past three months had been “hell,” describing the exponential growth in model capabilities as astonishing—and accelerating unabated. He warned that organizations have little time left to prepare for potential AI-driven cyberattacks, and those who dismiss security risks or claim systems are infallible should not remain in leadership roles. He urged greater attention to those consistently raising alarms and identifying flaws.
Pressure is also mounting in Washington. Rapid expansion of data centers has sparked public backlash over rising electricity costs, and the midterm elections loom. Within the White House, some senior officials express caution about AI models, fearing economic disruption. Yet Trump continues to resist slowing AI development, arguing that the U.S. must keep pace with China and that technological prosperity remains vital to stock market performance. On his social platform, he dismissed fears of AI taking over the world or causing human extinction as “a hoax.”
Insiders note a recent surge in apocalyptic rhetoric, which paradoxically appears to strengthen Trump’s and a small circle of advisers’ suspicions—some believe these narratives are being amplified by China. Trump’s allies have also criticized tech executives for simultaneously warning the public about existential risks while complicating policy decisions. One adviser directly labeled them “idiots.” Others have taken issue with Dario Amodei’s call for a slowdown, arguing it damages public perception. Yet after the meeting, Trump publicly praised Amodei as “fantastic.”
In sum, Trump seeks to loosen constraints on AI development through rebranding and self-regulation. But safety incidents and public unease remain unresolved. Whether the agreement can effectively constrain these major players remains uncertain. Should a significant failure occur in the future, responsibility may still ultimately fall back on the White House.
Xi Jinping's Visit to the United States
Original article: toutiao.com/article/1877739138090052/
Disclaimer: The views expressed in this article are those of the author and do not necessarily reflect the position of any official entity.