By Benjamin Guggenheim

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

• Anthropic says it’s releasing its most advanced model, Claude Fable 5, back to the public after the Commerce Department lifted its export directive. The process surrounding government AI intervention is as opaque as ever.
• An Intel shareholder has filed a suit against the Intel board and the Commerce Department over the U.S. government’s 10 percent stake in the company.
• The case exposes new details on Intel and Commerce’s negotiations. It also shows the legal and political risks facing the quantum and semiconductor companies that have partnered with the administration.

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

Anthropic’s Claude Fable 5 is back after roughly 18 days of suspension. What exactly happened, and where do we go from here?

Inside Commerce’s decision: According to the Tuesday letter I’ve obtained that is addressed from Commerce Secretary Howard Lutnick to Anthropic Chief Compute Officer Tom Brown, Lutnick says export controls on both Mythos 5 (the restricted model for a select number of business partners) and Fable 5 (for consumer use) have been lifted in light of evaluations by the Commerce Department’s Bureau of Industry and Security of the model’s “diversion risks.”

Lutnick says that Anthropic has committed to addressing security risks of future models, but that Commerce reserves the right to impose a license requirement.

What Anthropic says: In Anthropic’s blog post published Tuesday, the company says Fable 5 will be available for consumers again on Wednesday. Importantly, by way of explaining the sequence of events that led to the decision, Anthropic does not concede that the export directive was justified.

The company writes that the so-called “jailbreak” that triggered the directive (and which was first identified by Amazon researchers) was demonstrated also in less capable Claude and OpenAI models.

“Even so, we moved quickly to address the reported bypass,” Anthropic says.

(Amazon founder Jeff Bezos owns The Washington Post, and The Washington Post also has a content partnership with OpenAI.)

Government collaboration: Anthropic says that it is deepening collaboration with the U.S. government to provide the administration pre-deployment access for evaluations of AI models. Anthropic also pledges to “quickly investigate, triage and notify appropriate government counterparts” of significant jailbreaks and misuse patterns.

Importantly, Anthropic is partnering with Google, Microsoft, Amazon and other partners in its Glasswing program to come up with an industry consensus framework for determining the severity of jailbreaks (according to metrics such as capability gain, ease of weaponization and discoverability).

What it means: Despite the fact that we just witnessed a significant government intervention in AI, the process surrounding it is as opaque as ever. Legal experts have panned Lutnick’s original directive to Anthropic as confusingly written and legally dubious — so we don’t have a very clear idea of what export control authority the administration would cite if it decides to build a more transparent AI governance regime.

Anthropic’s explanation has also left cyber researchers skeptical that anything substantively changed to justify Fable’s retraction — and the model’s subsequent release by the U.S. government.

“Anthropic verifies that none of the jailbreaks provided a capability beyond what many other models, including Chinese models, could do,” cyber researcher Alex Stamos said in a post on X. “This is a huge own goal by the U.S., and we will see how bad U.S. models get over the next six months.”

The Claude Mythos logo is displayed on the screen of a smartphone. (Photo by Samuel Boivin/NurPhoto via Getty Images)

Looking forward: The optimistic view of the situation — as articulated by former White House AI adviser and OpenAI’s new Head of Strategic Futures Dean Ball on X — is that “this opacity does not lend itself well to a stable, investable, trustworthy industry over time … but we made progress today.”

Some takes are gloomier, and it’s easy to see why. While users can still use Fable to code, the much more stringent safeguards are more liable to block users if they’re querying something cyber-related and move the users down to a less-capable model.

The blunt instrument suggests that Anthropic expects heightened government scrutiny of future frontier models amid an ongoing debate over how to measure security risks.

And it points to a problem that lawmakers on the House Science Committee have been thinking about: The Commerce Department doesn’t currently have the capabilities, funding, or staff to keep up with rapidly progressing AI models or the assessments of frontier labs’ own researchers.

Some prominent industry players, such as cloud company CEO Matt Calkins, tell me that they can see this disparity ultimately leading to nationalization for the frontier labs.

Government equity in emerging tech

The Trump administration’s unprecedented practice of taking equity stakes in large companies is starting to face legal challenges. The challenges are bound to have ripple effects in the semiconductor, quantum and AI supply chain industries.

As Lead Global Security Analyst Josh Rogin outlines in a new WP Intelligence report, a suit by Intel shareholder Richard Paisner against the Commerce Department and Intel’s board of directors has exposed new details on how the U.S. government’s 9.9 percent equity stake in the company was brokered.

The case alleges that the deal breached the Intel board’s fiduciary duties to its shareholders. And it’s poised to set a major precedent that will determine whether the administration’s other equity deals remain intact.

Here’s what we know: The plaintiff alleges, according to board documents produced for Paisner’s attorneys under Delaware’s books-and-records statute, that Chips Program Office General Counsel Dave Shapiro told Intel CFO David Zinsner last Aug. 18 that Intel should not “make the mistake of thinking that this is a negotiation.”

Shapiro allegedly warned Zinsner that a rejection of Trump’s equity demand would have unpredictable “consequences,” and that Intel should “consider benefits of having a friend in this administration.”

The board was allegedly notified of the deal Aug. 20 and approved the deal two days after — without legal counsel advising on whether the transaction was lawful.

Intel declined to make Zinsner available for comment. The Commerce Department did not respond to a request to make Shapiro available for comment.

“The only special interest guiding the Trump administration’s decision-making is the best interest of the American people,” White House Spokesman Kush Desai told WP Intelligence.

Both Intel and Commerce have filed motions to dismiss the case, which remained outstanding at the time of publication.

Industrial policy and political risk

A slew of equity deals: WP Intelligence found 20 equity deals that have given the U.S. government equity stakes worth more than $50 million. The Commerce Department holds stakes in semiconductor quantum computing and critical minerals companies such as Intel, PsiQuantum, IBM’s quantum subsidiary Anderon, GlobalFoundries and Diraq.

The Department of Defense holds stakes in MP Materials, Trilogy Metals, Vulcan Elements and L3Harris. The Energy Department holds a stake in Lithium Americas.

From a public policy standpoint, the administration makes the case that these stakes bolster national security and allow the U.S. government to coordinate supply chains to remain competitive with China. But legal experts say it also distorts the marketplace and incentivizes anticompetitive behavior.

Increased scrutiny: Josh writes that the equity deals negotiated by the administration are subject not only to legal challenges but also political risk as Democrats ramp up scrutiny of these deals.

For example, House Science Committee Ranking Member Zoe Lofgren (D-California) has criticized the USA Rare Earth deal given the company’s ties to Cantor Fitzgerald, which is run by Lutnick’s sons. The Vulcan Elements deal has also received heavy scrutiny since the company’s raised Series A financing from Donald Trump Jr.’s venture capital firm, 1789 Capital.

Trump Jr., 1789 Capital and Vulcan did not respond to requests for comment.

“The recently announced definitive agreements with the U.S. government followed extensive diligence conducted by technical professionals and advisers within the federal government,” a spokesperson for USA Rare Earth said.

The outlook: For companies contemplating U.S. government equity deals, Josh says the Intel case shows the importance of boards to document their decision-making process and properly consult shareholders.

The best way to ensure that the deals remain intact long-term is to advocate for congressional authorization and regulation of theses equity deals.

Democrats may well seek to renegotiate the deals or impose new conditions if a progressive administration takes power, Josh argues.

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.