Anthropic becomes a cautionary sovereign-AI fable

Alibaba Holding

Mon 13:00

The author is a Reuters Breakingviews columnist.  The opinions expressed are her own. Refiles to add and fix hyperlinks.
By Karen Kwok

    NEW YORK, June 15 (Reuters Breakingviews) - Anthropic’s most powerful chatbots are living up to their names. On Friday, the White House placed severe export controls on Mythos, a model purported to function as a hacking super-weapon, and Fable, a sanitized variant meant for public consumption. Boss Dario Amodei responded by suspending access altogether. His $965 billion company’s breakneck progress has opened up its own regulatory Pandora’s box. Global allies may take it as a cautionary tale about depending on the United States.
The dispute arose over concerns that safeguards preventing Fable’s use as a hacking tool can be bypassed. Anthropic says the problem is narrow and does not expose novel capabilities. Yet Andy Jassy, CEO of Amazon.com AMZN.O, an investor in Anthropic, felt the need to flag the danger to officials, Reuters reported.
It’s a messy example of how untamed and uncertain AI regulation is. Anthropic tried to stay a step ahead, releasing Mythos only to select partners to find cybersecurity holes. After practically touting his products as inherently dangerous, though, Amodei may have inadvertently set a new precedent: chatbots cannot simply be thrown out to the public anymore.
That even close allies like the United Kingdom could be cut off from silicon smarts raises further alarm. White House officials have said that U.S. AI dominance will be built on exporting American technology, making it the global standard. Revealing that these exports carry a regulatory kill-switch threatens this project.
The European Union has already pushed U.S. cloud giants to store and process data locally. Singapore and the United Arab Emirates are developing models better suited to their languages and needs. Such alternatives are not cheap: McKinsey estimates local AI offerings can cost 10% to 30% more. Washington has made that insurance premium easier to justify.
Complicating matters further, the underlying technologies that make Mythos and Fable possible pass through a host of global chokepoints, like Taiwanese semiconductor foundry Taiwan Semiconductor Manufacturing 2330.TW, or Dutch company ASML's ASML.AS lithography machines. A world of rising export controls could make advanced AI models harder or more expensive for everyone to build.
Open-source alternatives that run on a user’s own hardware and cannot be revoked will look yet more attractive. That’s a leg-up to China’s industry, which has focused on cutting-edge open models like DeepSeek and Alibaba's 9988.HK Qwen.
More prosaically, as Anthropic heads towards an initial public offering, it must presumably explain to investors that the White House can quarantine the models it burns billions of dollars to train. How they can gauge this danger, unbounded by an independent regulator or explicit legislation, is unclear. This specific situation may well be resolved with quick fixes. Broader concerns cannot be stuffed back into the box.
Follow Karen Kwok on LinkedIn and X.
     
CONTEXT NEWS
Anthropic said on June 12 that the Trump administration had ordered the company to block any foreign nationals, whether inside or outside the United States, from using its latest and most advanced AI models, Fable 5 and Mythos 5. In response, Anthropic said it would disable access to the models globally.
In the blog post, Anthropic said the U.S. government told the company it believes there is a method of bypassing, or "jailbreaking," a safeguard ​against using the model to find cybersecurity holes. The bypass found only "minor" security flaws that other publicly available models can also find, Anthropic said. 
Andy Jassy, CEO of Anthropic investor Amazon.com, was ​among those who raised concerns to senior Trump administration officials this week, Reuters reported on June 13, citing a person familiar with the matter.
"As a leading cloud provider that serves a large number of private and public sector customers, it’s not uncommon for governments to seek ​our counsel on potential security risks," an Amazon spokesperson said. "When they occur, we don’t share the details of these discussions."

(Editing by Jonathan Guilford; Production by Aditya Srivastav)
((For previous columns by the author, Reuters customers can click on KWOK/karen.kwok@thomsonreuters.com))