By Benjamin Guggenheim

Not a subscriber? Sign up here to get this newsletter in your inbox.

• The Trump administration’s export control directive on Anthropic’s Fable model has had a chilling effect on the AI industry. Chinese AI leaders are seizing on the uncertainty, arguing that their cheaper and more accessible models will win out in the longterm.
• The House Science Committee advanced a suite of bills Thursday to establish new AI workforce training programs and beef up the Commerce Department’s Center for AI Security and Innovation. The real battle is being fought over children’s online safety legislation.
• A run-down of the stories you may have missed this week, including coverage of the administration’s new quantum executive orders and New York Assembly member Alex Bores’s loss in the big AI primary.

This is Benjamin Guggenheim. Welcome to WP Intelligence’s AI & Tech Brief, where we examine the transformative technology of artificial intelligence at the intersection of innovation, policy and power. Get in touch at Benjamin.Guggenheim@washpost.com or on Signal at ben_guggenheim.88.

The Lead Brief

The Trump administration’s export control directive ,which pressured Anthropic into retracting its Fable model, has had an undeniable chilling effect on the AI industry — which is now under the impression that the government can capriciously shut down AI systems at any moment it wishes.

And while frontier labs can technically keep developing their models during this period of uncertainty, AI industry experts such as Hammond are arguing that the Fable directive has helped China catch up in the AI race.

That’s the case because of several factors: 1) frontier labs don’t get important feedback and information from testing their models at scale out in the field; 2) enterprises start looking at open-source models that are widely accessible and not at risk of being unplugged by the U.S. government; and 3) there’s uncertainty over whether non-U.S. nationals, including employees of Anthropic and OpenAI, can continue to work on the models under the export control directive.

“The lack of a real standard or standard-making process is already driving the foundation labs to be extremely careful about refusals in their current models,” said cybersecurity expert Alex Stamos on X. (Stamos led an open letter to Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross on June 14 asking them to lift the export control directive on Fable.)

“This will drive U.S. security companies and researchers to use post-trained Chinese as their go-to primaries,” Stamos said, adding that he is aware of a number of companies that have started using Chinese open-source model as backups in the event of another unpredictable shutdown.

The Trump administration has now also asked OpenAI to delay the release of its most advanced model, GPT-5.6, as the Information first reported. The publication reported that the administration would approve the release of a limited preview form to a handful of enterprise customers on a case-by-case basis.

The China angle: The regulatory uncertainty in the U.S. is a boon to the Chinese AI companies and leaders, who are betting that “good enough” models operated at cheaper rates will win the AI adoption race, Rebecca Tan and Lyric Li report for The Post.

Many enterprises need “a fraction of the capability” of the most advanced U.S. AI models, according to a salesperson at a leading Chinese AI firm, who spoke on the condition of anonymity because they were not authorized to speak to the press.

“If we can provide 80 percent of the value at a much lower cost, that’s enough. We’ll take that market share,” the salesperson told The Post.

At a forum in Paris last week, Alibaba’s chairman, Joseph Tsai, pointed to the administration’s recent directive for Anthropic in his pitch for Europeans to consider using the company’s “Qwen” class of large language models.

“The problem is, right now, all your eggs are in one basket,” Tsai said. “Why not choose a second basket and diversify your eggs?”

Joseph Tsai, chairman of Alibaba Group (AP Photo/Ng Han Guan, File)

Elicit, alleged espionage: Meanwhile, in a new letter to two U.S. senators, Anthropic accused Alibaba of using so-called “distillation” attacks to extract and steal Claude’s capabilities. (Alibaba did not respond to a request for comment from media outlets.)

These kinds of distillations have been of growing concern to lawmakers on the House Foreign Affairs Committee, which advanced a suite of bills in April that would clamp down on the practice.

What it means: The environment in Washington of chaos and uncertainty will incentivize hyperscalers and frontier labs to more aggressively push for a clear AI regulatory framework.

The uncertainty has also shifted the political incentives around AI regulation. The pro-innovation camp in the Trump administration has historically argued that a “hands-off” approach was the best way to compete with China.

But, after the intervention into Anthropic’s Fable model, the only way to resolve business uncertainty will be to establish a clear framework for AI governance.

Case in point: Google put forward its own white paper on AI governance on Thursday, calling on Congress to establish an independent regulatory body that could define and verify models’ safety and security benchmarks prior to their release.

Google proposes that the body pay “market-rate” salaries to attract staff with significant AI expertise in the public sector.

The Legislative Chessboard

The House Science Committee advanced a suite of long-anticipated AI legislation Thursday.

The roster included legislation that would:

• Expand AI and workforce training programs under the National Science Foundation.
• Direct the National Institute of Standards and Technology (NIST) to improve the measurement and analysis of water use by data centers.
• Establish a National Artificial Intelligence Research Resource to democratize access to AI resources for small businesses and academic researchers.
• Officially codify and authorize $20 million a year for the Commerce Department’s Center for AI Security and Innovation (CAISI) under NIST. Under this bill, CAISI is directed to set up voluntary agreements with frontier labs to evaluate models for national security risks — especially in the cyber, chemical and biological domains. CAISI would also track how U.S. models compare to those of foreign adversaries.

What it means: Most of the bills were advanced unanimously by the House Science Committee and are relatively uncontroversial. The bigger political battle standing in the way of an AI package is being fought in negotiations over children’s online safety legislation.

The House is expected to put a package of bills with social media and chatbot guardrails for minors on the floor next week. But influential lawmakers, such as the top Democrat on the Senate Commerce Committee, Sen. Maria Cantwell (D-Washington), are coming out vocally against the House package, arguing that it significantly weakens the proposals in the Senate’s “Kids Online Safety Act.”

The House legislation “narrow[s] the platforms that are covered, leaving online gaming services out,” Cantwell said in a press call with reporters on Friday. The House bill also “completely [ignores] major problems like eating [and] social disorders, suicide behavior, addiction, harassment, or diagnosed depression or anxiety.”

We are “concerned that the House of Representatives would take a weakened approach and gut much of that legislation in its ability to actually enforce the law,” Cantwell said.

ICYMI

• My Monday edition covered the Trump administration’s release of two executive orders focused on boosting quantum technology. The House has put forward its own proposal on children’s online safety legislation as AI negotiations gain steam in the Senate.
• My Tuesday edition unpacked the administration’s new quantum industrial policy. I interviewed John Levy, CEO of quantum systems company SEEQC, about how the EO sent an important signal on to the capital markets — and the importance of the technology ecosystem that enables quantum.
• My Wednesday edition analyzed New York Assembly member Alex Bores’s loss in the “big AI primary” in District 12. I also looked at why tech stocks in the memory and storage space took a tumble.
• My Thursday edition covered how lawmakers in Canada and the U.K. are supporting campaigns to prevent the development of “AI superintelligence.” ControlAI U.S. Executive Director Connor Leahy has been briefing members of Congress on the risks posed by an AI system that is smarter than humans.

This newsletter is published by WP Intelligence, The Washington Post’s subscription service for professionals that provides business, policy and thought leaders with actionable insights. WP Intelligence operates independently from The Washington Post newsroom. Learn more about WP Intelligence.