AI firms pledge safety measures and mutual oversight

No penalties for accidents or violations of the agreement

Pact closely mirrors Biden-era voluntary commitments from 2023

Critics say companies are effectively setting their own safety standards

Trump also pushes expansion of data center investment

US President Donald Trump speaks to reporters at the White House in Washington after meeting with executives from major AI companies on Wednesday. [AFP]
US President Donald Trump speaks to reporters at the White House in Washington after meeting with executives from major AI companies on Wednesday. [AFP]

US President Donald Trump hailed a voluntary AI safety agreement signed by leading technology companies as "almost like a constitution" on Wednesday, but experts say the pact lacks any enforcement mechanism and amounts to little more than a symbolic gesture. Critics noted that the agreement leaves AI companies to set their own safety standards while imposing no consequences for accidents or violations — prompting some experts to say it can simply be ignored.

The deal also drew comparisons to a similar voluntary commitment that the Joe Biden administration secured from AI companies three years ago. The Biden administration later concluded that voluntary pledges alone were insufficient to address AI risks and established a government-level safety evaluation framework — one that Trump scrapped after taking office.

According to the BBC and The New York Times, Trump met Wednesday at the White House with the chief executives of the country's most powerful AI companies, including Elon Musk of SpaceX AI, Nvidia CEO Jensen Huang, Meta CEO Mark Zuckerberg, Google CEO Sundar Pichai, Anthropic CEO Dario Amodei and OpenAI President Greg Brockman.

Following the meeting, the companies signed an agreement committing to build safeguards that ensure AI models behave as intended, and to rapidly detect and correct problems when they arise. The pact also calls for working with independent auditors to assess whether AI systems are functioning normally and for preventing AI from hacking or accessing other systems in unintended ways.

Trump said he signed the agreement himself in his capacity as president, describing it as "almost like a constitution in some ways." He also pointed to mutual oversight among companies as a safety mechanism, saying, "They're actually going to be watching each other."

The agreement, however, is a voluntary, self-regulatory arrangement grounded in moral commitment. It contains no specific provisions on how much liability companies would bear if an AI system caused harm, nor any mechanism for sanctioning companies that fail to honor their pledges.

Experts were quick to flag the omission. Kimberlee Weatherall, a law professor at the University of Sydney, told the BBC the agreement was "a very disappointing outcome," saying it effectively allows companies to "set the standards for safety themselves" without specifying any consequences for violations. "This agreement is a distraction and should be ignored," she said.

Jeanie Patterson, a researcher at the University of Melbourne's AI and Digital Ethics Centre, also said self-regulation alone risks leaving companies to do only the bare minimum on safety. Because the companies developing and commercializing AI are also the ones setting the safety standards, she argued, the arrangement is tantamount to letting players referee their own game.

The New York Times noted that the agreement closely resembles the voluntary safety commitments Biden secured from major AI companies in July 2023.

At the time, the core idea was that because government regulation struggled to keep pace with rapidly advancing AI models, companies should build their own oversight systems and layer multiple safeguards against malicious use of the technology.

But as cases of AI models behaving in unexpected ways continued to mount, debate grew over whether corporate self-regulation was sufficient.

Ben Buchanan, a professor at Johns Hopkins University who helped design the 2023 voluntary commitments, told The New York Times: "We knew in 2023 that voluntary commitments alone weren't enough — that's why President Biden later issued an executive order."

The Biden administration subsequently used that executive order on AI safety to establish the US AI Safety Institute within the Commerce Department. The institute was tasked with developing safety standards for new AI models and testing them before release, supplementing corporate self-regulation with government-level verification.

Trump, however, revoked Biden's AI executive order after taking office, arguing that AI regulation could slow the pace of technological development and put the United States at a disadvantage in its competition with China.

At Wednesday's meeting, Trump again signaled that his priority is expanding investment in AI rather than tightening regulation of the industry. He reaffirmed his administration's support for the construction of AI data centers in particular.

In the United States, community opposition to data centers has been growing over concerns about their enormous power consumption, rising electricity bills and environmental impact. A recent Marist poll found that 65 percent of registered voters oppose having a data center built in their community.

Trump said companies could offer fiscal and educational support to communities that host data centers, adding that "these big, powerful and very wealthy companies will make an incredible contribution to the communities."

On the same day, Trump also signed an executive order directing the federal government to use the term "Super Intelligence" — or SI — in official documents and websites in place of "artificial intelligence."

Wednesday's meeting brought into sharper focus the direction of the Trump administration's AI policy: minimize mandatory government regulation, entrust AI safety to corporate self-regulation, and at the same time accelerate investment in AI infrastructure such as data centers.


sjy@heraldcorp.com