Dan Gallagher and Asa Fitch
Anthropic has made a habit of sounding warnings about AI's dangers. But Anthropic itself is creating some of the world's most powerful and potentially dangerous models.
That makes it harder for the company to combat moves like the Trump administration's order on Friday barring all foreigners -- including Anthropic's own employees -- from accessing its Mythos 5 and Fable 5 models.
Anthropic can point to the guardrails it has placed on its models. It can also demand equal treatment with competitors. But its own communications make it hard to argue that its models aren't inherently dangerous.
It may be the case that its warnings have merit. In a blog post earlier this month, the company suggested a global pause in development if self-improving AI systems take off. That wouldn't be completely unreasonable, even if it would be hard to enforce.
It also is wise for a developer like Anthropic to take the potential danger of what it is doing into account as it develops cutting-edge AI tools. Even from a self-interest perspective, preventing harms that lead to scrutiny and overregulation is logical.
It is less clear whether Anthropic benefits by being so loud about it all. Anthropic is now the world's highest valued private company, with its latest post-money valuation hitting $965 billion and topping OpenAI's $852 billion valuation, according to data from PitchBook.
Such a vaunted valuation -- which could easily top $1 trillion once Anthropic gets its own pending IPO out the door -- means the company can't afford hiccups that stunt its growth potential.
And the latest row is the second major dispute with the Trump administration, which deemed Anthropic a supply-chain risk over its refusal to allow the unrestrained use of its software by the military. Anthropic sued in March to undo that designation. The case is ongoing.
Anthropic's success at developing cutting-edge AI models and corporate-focused tools has given it a strong edge, despite the ChatGPT developer's first-mover advantage. Mere announcements of new Anthropic tools have triggered selloffs in stocks of software and cybersecurity companies thought to be in its competitive crosshairs.
The rollout of Mythos stood out even more. Anthropic introduced the model to the world with the claim that it was too dangerous to release in full. The world got the message -- maybe a bit too well.
This is an edition of the WSJ AI & Business newsletter, a weekly digest to help you make sense of AI's impact on business with news, insights and data from our global team of technology journalists. If you're not subscribed, sign up here.
SpaceX Goes for Cursor
Less than three trading days after its listing, SpaceX is using its stock to buy Anysphere, the parent of the coding-agent company Cursor. The $60 billion deal would bring Elon Musk's company a competitor to Anthropic's Claude and OpenAI's Codex, and would boost revenue in SpaceX's AI business. But it will do little to stem operational losses in the division, which totaled about $6.4 billion in 2025. Investors didn't seem to mind: SpaceX's shares climbed more than 12% in early trading Tuesday -- raising the company's market cap past Amazon's.
The Number
The amount raised by Chinese AI startup DeepSeek in its latest funding round. Investors are now valuing the company at more than $50 billion. Founder Liang Wenfeng, who held nearly 90% of DeepSeek before the financing round, made the largest contribution to the fundraise, sources told The Wall Street Journal.
What the Humans Are Saying
AI in Charts
It's far from certain that Qualcomm can become a major force in the data-center chip business. But unlike other speculative AI chip plays, investors aren't placing a crazy bet that it will be.
Qualcomm's stock price has risen strongly recently, but it is still valued below many of its peers on a forward price-to-earnings basis. That is partly because the company remains closely associated with chips for smartphones, which haven't had a good run: smartphone sales are expected to fall this year because of difficulties obtaining memory to make them.
But Qualcomm does a lot of other things these days. It has built up large automotive-chip and internet-of-things chip businesses. It is making chips for personal computers. And more recently, the company is making a play for AI data center chips.
That move isn't assured to succeed, but Qualcomm has reasonable chances if the AI boom doesn't wane and computing demand remains red-hot.
AI in the Wild
Three years into training to be a court reporter, Jayne Williams is used to warnings that AI will soon take over her budding career. The profession has become an example of AI's limitations in replacing human skill in the real world. In an actual courtroom, court reporters record nonverbal cues like gestures and transcribe through distracting courtroom noises like coughs or door slams. Other times, they must gently ask witnesses to repeat themselves while recounting traumatic testimony.
Other Highlights From the Week in AI
-- A consumer is suing Anthropic and seeking class-action status, claiming
that the company was misleading about limits on its premium plans.
-- OpenAI is getting legal pushback, too -- from a coalition of attorneys
general that is investigating its impact on a range of people.
-- Elon Musk's xAI, which is now part of SpaceX, had a case dismissed in
California in which it claimed its rival OpenAI stole trade secrets.
-- Databricks is releasing AI agents that help professionals get answers
from their business data, aiming to expand beyond its core data
offerings.
About Us
WSJ AI & Business is a weekly look at AI's transformation of the business world. This newsletter was curated and edited by Dan Gallagher and Asa Fitch. Reach them at dan.gallagher@wsj.com and asa.fitch@wsj.com (if you're reading this in your inbox, you can just hit reply). Got a tip for us? Here's how to submit.
(END) Dow Jones Newswires
June 16, 2026 12:05 ET (16:05 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments