4 hours ago
4
AI + ML
Claude maker warns authoritarian regimes could set the rules unless Washington tightens chip and model controls
AI monger Anthropic wants America and its allies to tighten measures aimed at curbing China's AI progress, warning of the consequences if "authoritarian governments" take the lead rather than Uncle Sam.
In a lengthy missive posted on its website, the San Francisco-based org says it expects AI to deliver "transformational economic and societal impacts" in the coming years, and whether the transition goes well depends on where the most capable systems are built first.
Since the technology is advancing swiftly, democratic countries have only a limited time in which to act, Anthropic believes. The measures it wants to see are nothing new: enforcing tighter export controls on chips used for AI development, such as Nvidia's GPUs, and cutting off access to American AI models.
Recent history suggests these controls "have been incredibly successful," it says. But if Chinese researchers are only several months behind the US in AI capabilities, as many experts estimate, how successful can those efforts have been?
AI labs in China have only built models that come close to those in America because of their talent and their knack for exploiting loopholes to get around export controls, Anthropic claims, along with distillation attacks that "illicitly extract the innovations of American companies."
Many will suspect this is Anthropic's chief motivation in calling for action against China. Back in February, the Claude model maker accused China-based rivals including DeepSeek of using distillation to train their models by siphoning knowledge from Anthropic's own.
As The Register pointed out at the time, accusing China of copying, while using content created by others to train your own models, shows a staggering lack of self-awareness from the AI industry.
Anthropic's sermon also shows blinkered thinking. It implies that China can only advance by riding on America's coattails, and is incapable of innovating. This is despite the shockwaves generated by the release of the DeepSeek R1 model early in 2025, believed to be on a par with the best US models.
Numerous reports also indicate that Chinese organizations have made huge strides with domestically developed AI silicon, and Beijing even tried to discourage tech companies in the country from buying and using Nvidia chips.
Anthropic sets out two scenarios for what the world could look like in 2028, a date when it expects "transformative AI systems" to have emerged.
In the first scenario, America has "successfully defended its compute advantage," and "democracies set the rules and norms around AI." The second has China overtaking the US, leading to AI norms and rules being shaped by authoritarian regimes, with the best models enabling "automated repression at scale."
Another problem with Anthropic's plan is that many countries, especially in Europe, view both American and Chinese AI supremacy as a threat to democracy. There is a concerted push in Europe for "digital sovereignty" to minimize reliance on US technology, for example. Others warn it could erode democracy in America itself.
Anthropic can draw little comfort from the Trump administration, which has a constantly shifting attitude to China. Export controls were said not to be high on the agenda during the President's trip to Beijing this week, and it was reported that the US has now cleared around 10 Chinese firms to buy Nvidia's second-most powerful AI chip, the H200. ®
English (US) ·