Secretary Hegseth And Chairman Caine Hold Pentagon Press Briefing On Iran Strikes

(Alex Wong/Getty Images)

A blow-by-blow catch-up on how Trump vs. Anthropic has unfolded

A consequential week punctuated by war highlights the deep rifts that exist between Silicon Valley and Washington.

3/2/26 12:50PM

Last week’s tense standoff between Anthropic and Defense Secretary Pete Hegseth focused on the issue of how the startup’s AI could be hypothetically used to surveil Americans or kill people. By the weekend, that hypothetical scenario became very real.

Tensions started after a tense meeting at the Pentagon, which resulted in Hegseth giving Anthropic CEO Dario Amodei an ultimatum: he had until Friday at 5:01 p.m. ET to get on board with the Pentagon’s demands for full use of Claude for any legal purpose, or face the consequences. Here’s a quick timeline of how this fast-moving story unfolded:

Thursday 5:36 p.m. ET - Anthropic posted a statement from Amodei that the company will not acquiesce to Pentagon demands to allow the use of its Claude tool for any legal purpose.
Friday 3:47 p.m. ET - President Trump lashed out at the startup, calling it a “RADICAL LEFT, WOKE COMPANY,” and said he was directing all federal agencies to cease use of Anthropic’s software.
Friday 5:14 p.m. ET - In a post on X, Hegseth chided Anthropic’s refusal, and in an extraordinary move, said he would move to designate Anthropic “a Supply-Chain Risk to National Security,” effectively blacklisting the company’s tech from use by the US military or any contractors (like Palantir) working with the Pentagon on national security applications.
Friday 8:24 p.m. ET - In another blog post, Amodei defended Anthropic’s position and signaled that it may pursue legal action, calling the move “legally unsound.”
Friday 9:56 p.m. ET - OpenAI CEO Sam Altman posted on X that his company reached an agreement to deploy its models on the Pentagon’s classified network.
Saturday 1:15 a.m. ET - The American and Israeli attack on Iran commenced.
Saturday 8:21 p.m. ET - The Wall Street Journal reported that Anthropic’s Claude was used by the Pentagon to assess intelligence, identify targets, and simulate battle scenarios.
Sunday - Anthropic’s Claude app rose to the No. 1 spot on the iOS app store, dethroning ChatGPT.

The feud sent a chill through Silicon Valley, as companies wondered what control they would have going forward if the government decided it needed their tech.

<10%

Rani Molla4h

Despite a massive surge in corporate AI spending, the technology is broadly failing to deliver the massive cost reductions executives had anticipated, according to a new global survey from Bain & Co. shared with Bloomberg. The largest share of major companies measuring their AI returns — 40% — realized cost savings of 10% or less, with poor access to internal data cited as the primary roadblock. Most had expected higher returns. More concerningly, Bain warned that many companies are using their original, overly optimistic projections — rather than their actual savings — to justify funding their next wave of expensive AI investments, creating a “circular bet with a structural leak.”

Rani Molla

Meta thinks it can save its hardware division with AI subscriptions. There are reasons to be skeptical.

An $8 AI subscription won’t plug a $19 billion hardware hole.

A visitor is seen trying a Ray-Ban Meta glasses at a Meta

Jon Keegan4h

Anthropic confidentially files for IPO

Anthropic has filed confidentially with the Securities and Exchange Commission for its initial public offering. The IPO is expected to be one of the largest in US history, and will likely be joined by OpenAI, which is also expected to go public before the end of the year.

The company filed a draft S-1 form with the SEC, which does not indicate the price of the offering. The official public S-1, which will come later, will give potential shareholders a first look at the finances of Anthropic, which just last week announced that it raised $65 billion, reaching a valuation of $965 billion. This puts the company well ahead of archrival OpenAI, which is currently valued at $850 billion.

Anthropic confidentially submits draft S-1 to the SEC

Rani Molla6h

Prosus may thwart Uber’s bid for Delivery Hero

Uber’s aggressive pursuit of Delivery Hero could hit a major roadblock. After the European food delivery giant rejected Uber’s initial $11.6 billion buyout offer, the American company pivoted, scooping up a 37% stake in the open market.

Now, Prosus, formerly Delivery Hero’s largest shareholder, is plotting a counteroffensive.

Thanks to an EU regulatory waiver Monday that temporarily pauses its mandatory stock sell-down, the Amsterdam-based investment firm is reportedly looking to either increase its stake or rally other shareholders against Uber. The goal: block the takeover entirely or force a significantly higher premium.

Prosus has warned about the loss of European tech relevance if a US giant swallows the company. Meanwhile, investors are loving the drama: the takeover tug-of-war, which also includes DoorDash, has sent Delivery Hero stock soaring over 75% in the past month.

Uber faces Prosus challenge in Delivery Hero pursuit

Rani Molla7h

Tesla sales surge in European markets in May

Tesla sales surged across Europe in May, Reuters reports, with sales jumping double and even triple digits in a number of early-reporting markets. Of course, 2025 was a very difficult year for Tesla sales in Europe, so the growth is coming off notably small denominators.

Interestingly, the resurgence is happening without EU approval for supervised Full Self-Driving, something CEO Elon Musk predicted would cause sales to “improve significantly” after blaming the absence of the tech for its weak sales.

The company has received approval for a version of its FSD tech in the Netherlands, as well as Lithuania and Estonia, and expects “EU-wide” permission in the second or third quarter.