A voluntary AI self-policing agreement has been signed by President Donald Trump and top tech industry leaders, in one of the most ambitious — and controversial — steps in regulating artificial intelligence. The agreement, published on Trump’s Truth Social , calls for internal and external controls by the companies themselves, without binding legislation. The question is whether self-regulation is enough for a technology that is fundamentally changing the world.
At the White House, Trump and House Speaker Mike Johnson met with the heads of the largest artificial intelligence companies, which are investing hundreds of billions of dollars in developing AI infrastructure, including data centers in communities across the country. The agreement was signed by Trump, Anthropic CEO Dario Amodei , Google CEO Sundar Pichai , Meta CEO Mark Zuckerberg , OpenAI president Greg Brockman , Nvidia CEO Jensen Huang , and Elon Musk , founder of xAI , which is now part of SpaceX .
“ I see tremendous self-regulation. And they understand that they need to self-regulate ,” Trump told reporters outside the West Wing. The agreement opens the door to future legislation , but focuses on four voluntary steps that companies are required to take: implementing strong internal controls , working with an independent external auditor , establishing a committee on each company’s board of directors , and evaluating audit reports . “ Over time, it may make sense to codify these steps into laws and regulations ,” the agreement says.
See also: Donald Trump and Xi Jinping discussed AI guardrails and chips
What does the AI self-policing agreement include?
The AI self-policing agreement is not legally binding, but Trump called it “morally binding.” He also proposed appointing about 10 people to a committee that would “oversee the whole thing,” and said he would appoint someone to oversee the agreement in the coming days, after consulting with the industry. It’s worth noting that some of the actions outlined inthe agreement are measures that companies are already implementing or have committed to implementing.
Shri Narayanan , a professor of electrical and computer engineering at the University of Southern California , said the agreement's intent balances the freedom needed for innovation with regulation. He noted that innovation is not limited to improving the capabilities of AI models, but also extends to efforts to improve the security of the models.
But other experts are skeptical about whether self-regulation is sufficient. Robin Jia, an associate professor of computer science at the University of Southern California, said it’s critical to ensure that the intentions of leading AI labs are “well aligned with the broader interests of the American public.” He pointed to recent revelations of hacking and said, “The better the models are made, the greater the consequences of such incidents.” That observation highlights a critical security risk: more powerful AI models can be used as tools for more sophisticated cyberattacks.
Criticism and concerns about AI self-policing
Alex Pascal, executive director of the Berkman Klein Center for Internet and Society, has been highly critical of the agreement, calling the language vague and general. “The only way to reduce the risks and harms of AI for Americans is through strong legal accountability, regulation, and fundamentally changing the competitive dynamics that lead to unsafe decisions by AI developers,” he said. The criticism reflects a broader concern: that voluntary self-regulation cannot replace binding legislation, especially in a sector where the economic incentives for rapid development are enormous.
Dario Amodei of Anthropic, who has been one of the loudest voices calling for a slowdown in AI development, said: “The technology has very real risks. And the mechanism, how we address those risks, is still being debated.” This statement is revealing: even one of the signatories to the agreement acknowledges that there is still no clear answer on how the risks will be addressed.
See also: Donald Trump: China will not have access to top Nvidia AI chips
Meanwhile, OpenAI CEO Sam Altman said in a TV interview that the company is increasingly focused on security . The previous day, OpenAI had halted the release of a new model due to security concerns. “ We are adjusting our progress, which sometimes includes not training a model ,” he told CNBC. “ We have to do it in a way that we can make very confident statements about security so that people don’t have anxiety .”
America.gov: The new government AI chatbot
At the same event, Trump unveiled a government AI chatbot, America.gov, designed to help Americans navigate government services. “I will never stifle the development of a technology that will be bigger than the industrial revolution,” Trump said.
Secretary of State Marco Rubio pointed out that the chatbot on America.gov could soon be used to submit a passport application completely online, without the required visit to a processing center.
Trump also portrayed the global AI competition as primarily a confrontation between the US and China, stressing that there is no “second place.” The two countries agreed last week to create a communication channel to handle AI-related incidents during Chinese President Xi Jinping to Washington. However, Trump stressed that he does not want too much cooperation with China: “I didn’t want to do anything, because we are ahead. Why would I want to do anything?”
🔒 Protect your privacy with Proton VPN
Swiss VPN from the creators of Proton Mail — strict no-logs policy, strong encryption, and built-in NetShield that blocks ads, trackers, & malware.
- ✔ No-logs, based in Switzerland (except 14-Eyes)
- ✔ NetShield: blocks ads, trackers & malicious domains
- ✔ Covers all devices — free version available
The link is an affiliate link — SecNews may receive a commission at no additional cost to you. It does not affect the independence of our article writing.
On data centers, Trump said companies have agreed to address bipartisan opposition to building data centers in local communities by providing more financial support to local schools or taking steps to reduce energy costs. “You’ll see, data centers are going to be very popular,” he said. If the backlash continues, he added, “they’re going to be forced to go overseas or to other locations.”
See also: Kamala Harris appears instead of Trump in Google Search, Elon Musk reacts
Overall, the AI self-policing agreement reflects a broader trend: governments around the world are looking for ways to regulate AI without stifling innovation. The European Union has already enacted the AI Act, one of the strictest regulations in the world, while the US seems to prefer a more flexible approach based on self-regulation. The question remains: is the goodwill of companies enough, or is binding legislation needed to ensure the safety of a technology that is evolving at an unprecedented speed?
As SecurityWeek reports, this agreement is a significant milestone in the debate over AI governance, but many experts remain skeptical about its effectiveness. Cybersecurity and the security of AI models are intertwined: as models become more powerful, so does the potential for them to be misused by malicious actors. Self-regulation may be a first step, but society will judge whether it proves sufficient.
