Imperative to keep door open for dialogue: China Daily editorial

A $200,000 fine is not usually geopolitical drama. Yet for the US Department of the Treasury, it is.

In its first civil penalty under the Outbound Investment Security Program, the Treasury fined Amidi, LLC for failing to file a required notification about an investment by its controlled foreign entity in a Chinese embodied AI company. Amidi, the parent of Plug and Play Tech Center, was fined more than twice what its Chinese subsidiary put into a Shanghai robotics AI firm. The offense was not investing in the company, but failing to tell the US government about it.

The OISP requires approval being obtained for investments by US entities into certain activities in the quantum computing, artificial intelligence and semiconductor sectors.

In AI and robotics, even a $92,000 bet can now carry geopolitical weight. The irony is that these are precisely the industries in which separating the world's two largest economies makes the least economic sense.

The Treasury's penalty was levied in July and announced only on Wednesday, after China and the US agreed on the importance of dialogue on AI. What looked like a modest compliance matter has become a signal to every US investor wondering how far the new technological borders between China and the US have moved.

The case underscores how difficult it has become for US and Chinese capital to flow into each other's strategic technology sectors. The result is something more consequential than investment screening. It is the gradual construction of parallel, but still interacting, technological ecosystems.

That is an awkward proposition in robotics. AI may be trained in one country, embedded in hardware manufactured in another and deployed somewhere else entirely. China has built resilient industry and supply chains for batteries, motors, sensors, actuators and other key components. The US retains major strengths in frontier models, computing infrastructure and software. But the "brain" and the "body" of intelligent machines are difficult to separate. Trying to do so by decree risks making both more expensive and less innovative.

Nor is AI merely another industry. Its risks are escaping the categories that regulators have traditionally understood. Investment reviews can determine who owns a company. Export controls can determine who gets a chip. Antitrust rules can determine whether a merger is permissible. But none of these tools answers what happens when increasingly autonomous AI systems make decisions that neither their creators nor regulators anticipate.

The US' recent White House gathering of leading AI executives illustrates the problem. The companies agreed to a voluntary framework involving internal controls, external audits, board oversight and safety standards.

That is useful, but it also exposes the limits of relying on businesses to regulate a technology from which they stand to gain enormously. AI companies have powerful incentives to demonstrate safety, but equally powerful incentives not to fall behind their rivals. "Trust us" is therefore unlikely to be a sufficient regulatory philosophy.

The financial stakes are becoming larger, too. More than $1 trillion of global AI investment is expected this year, with the US accounting for a huge share. AI-related debt issuance has surged, while the technology's direct contribution to economic growth remains smaller than the capital being poured into it.

The danger is not that investment in AI is irrational. It is that enthusiasm can become a substitute for scrutiny. Capital markets have a familiar habit of extrapolating technological promise faster than society can assess its costs.

Late last month, China and the US agreed on the establishment of the China-US AI Dialogue to exchange views on risks and benefits related to AI. The next exchange will occur in November. The two sides also agreed to establish a bilateral communication channel for AI incidents.

Such mechanisms will not dissolve strategic competition. But competition without communication is an invitation to miscalculation.

AI is beginning to create systems unlike the corporations, states, social organizations and individuals for which modern regulatory frameworks were designed. The "monsters" of tomorrow may not be hostile machines; they may simply be systems whose autonomy exceeds the assumptions built into today's rules. Investment barriers cannot solve that problem. Nor can geopolitical competition.

China and the US, as the two major players in AI, therefore have a common responsibility beyond enhancing their own competitiveness. They must talk about safety even when they still disagree with each other on some important points. The sensible objective is not to eliminate competition or thwart each other's development, but to ensure that competition does not eliminate the capacity for cooperation and coordination. For an industry moving this quickly, keeping the door open for dialogue and exchanges may prove more important than building a higher wall.