Trump and Xi could discuss safeguards on AI

The United States and China seek safeguards to prevent their rivalry on artificial intelligence (AI) from escalating into a crisis, according to The Wall Street Journal, a week before the planned summit between the respective presidents of the two countries, Donald Trump and Xi Jinping.

The two giants are considering, according to well-informed sources, engaging in official discussions on AI, and possibly putting the subject on the agenda for Donald Trump’s visit to China, scheduled for May 14th and 15th.

Such an initiative would reflect “an awareness of the risk that the race of increasingly powerful AI models could trigger a crisis that neither of the two states would be able to manage,” writes the economic newspaper. “The Biden administration had started a dialogue with China, but its results were limited, and since then the risks have increased,” it recalls.

The possible outcomes of this dialogue remain unclear to observers, who mention, for example, a specific direct communication line dedicated to AI – considering that a telephone line for defense already exists and that “Beijing has often been reluctant to use it.”

“Common Enemy”

In the columns of The New York Times, Thomas Friedman has called for the creation of such safeguards “against malicious uses of AI, now that the latest models have demonstrated astonishing cyberattack capabilities.” The famous journalist refers to the Mythos model, which allegedly can detect security vulnerabilities that were previously undetectable. Such models could give small actors significant disruptive power. “China and the United States have a common enemy – and it’s not the Soviet Union,” the columnist asserts, referring to the two countries’ rapprochement in the 1970s against Moscow.

As anticipated by The Economist, the “Mythos moment” seems to have prompted a shift at the White House. Previously in favor of laissez-faire regarding AI, the Trump administration now considers establishing a control body for new AI models to prevent risks, as reported on May 4th by The Wall Street Journal. The Department of Commerce has entered agreements with Google DeepMind, Microsoft, and xAI to examine working versions of their new AI models before they are released to the public, as reported the next day by The Guardian.