Andrew here. A major national discussion right now is about President Trump’s recent financial disclosure, which notes an unprecedented $2 billion in earnings during his first year back in office.
Corporate leaders operate under strict compliance frameworks for their personal finances. But elected officials often face a different set of expectations — a regulatory gap I’ve frequently highlighted. Beyond the ongoing debate about the emoluments clause of the Constitution, the sheer scale of these figures may serve as a catalyst for meaningful policy overhauls.
Putting politics aside, the core concern is less about proving an explicit conflict of interest, and more about how even the appearance of one can erode public confidence in our democratic institutions.
Anthropic turns its top models back on
Fable and Mythos are back, mostly.
Anthropic will begin restoring access to its most powerful artificial intelligence models, after reaching an agreement with the Trump administration over safety guardrails.
The agreement concludes a blackout of the Anthropic models that lasted more than two weeks. But it’s unclear whether the move will settle growing concerns about the administration’s increasingly interventionist approach to regulating A.I. — and as rivals from China rapidly catch up.
What’s happening:
-
Anthropic will make Fable available to users starting on Wednesday.
-
Mythos will be available again for certain U.S.-based organizations, with Anthropic seeking to restore access to the broader set of domestic and overseas customers who were part of the company’s Glasswing program.
Anthropic agreed to additional safety guardrails for the models. Here’s what Commerce Secretary Howard Lutnick wrote in a letter to Anthropic lifting the ban:
Anthropic has agreed to proactively detect and address security risks associated with the models; to work diligently with the U.S. government on protocols and standards and releases for Mythos, Fable and future models; and to inform the U.S. government of any malicious activity.
More specifically, according to an Anthropic blog post, the company has addressed what Amazon and the administration said was a way to bypass Fable’s safeguards around cybersecurity.
(Anthropic cautioned that its models will have to scrutinize more potential false positives, which will be costly.)
Has the damage already been done? Even though the administration lifted the export restrictions on the Anthropic models, Lutnick wrote in his letter that his department “reserves the right” to reimpose those limits “should circumstances change.”
That may be a standard caveat. But it reinforces the growing suspicion in Silicon Valley and beyond that the Trump administration will play a stronger hand in regulating A.I., an industry it previously treated with a light touch.
-
“This opacity will not lend itself well to a stable, investable, trustworthy industry over time,” Dean Ball, a former A.I. adviser to the administration who now works at OpenAI, wrote on X. But, he added: “The status quo is not tenable, but we made progress today.”
And Chinese A.I. is still breathing down Silicon Valley’s neck. Tech experts are still talking about models like GLM-5.2 from Z.ai, which outperforms most Western equivalents, save the highest-end Anthropic and OpenAI ones. And it’s open-source and freely available.
“This was a huge own goal for the US, and we will see how bad US models get over the next six months and if Chinese models become noticeably better for cyber work,” Alex Stamos, a former chief security officer at Facebook, wrote on X.
HERE’S WHAT’S HAPPENING
The Trump administration reportedly plans to drop the U.S.-Mexico-Canada trade deal. The administration is expected to announce on Wednesday that it will not extend the agreement, known as U.S.M.C.A., starting a decade-long countdown to the end of a 32-year-old free-trade zone. Elsewhere, the E.U.’s side of a trade deal it struck last year with the U.S., which will eliminate import duties on many American products, comes into force on Wednesday.
An independent studio buys the Sam Altman movie that Amazon dropped. Neon purchased “Artificial,” a film about Sam Altman and his rocky road to becoming the C.E.O. of OpenAI, after Amazon put it up for sale (months after announcing a $50 billion investment in the artificial intelligence lab). Neon won best picture Oscars in 2020 and 2025 for “Parasite” and “Anora.”
Britain may intervene in the takeover of Warner Bros. Discovery. The country’s culture secretary, Lisa Nandy, said she was inclined to do so given her concerns about the concentration of media and news ownership. Paramount, the planned buyer of Warner Bros. Discovery, has offered remedies to the E.U. to resolve the bloc’s concerns about the transaction.
SCOTUS puts a data deal in jeopardy
The fallout from the Supreme Court’s chipping away at the independence of agencies like the Federal Trade Commission is rippling across the Atlantic Ocean, putting a major international treaty at risk.
The ruling threatens to disrupt a data privacy agreement between the United States and the European Union, and could have big implications for American businesses serving European customers, David Meyer reports.
The context: The F.T.C.’s role as an independent privacy regulator is a bedrock for the E.U.-U.S. Data Privacy Framework, an agreement that allows American firms to import Europeans’ personal data in compliance with E.U. data-protection laws. Adopted in 2023, the deal enables companies like Meta and Google to seamlessly serve users across Europe.
But now that the Supreme Court has undercut the F.T.C.’s independent authority, Max Schrems, the prominent Austrian lawyer and privacy activist, is calling on the European Commission to scrap the deal.
-
Schrems was responsible for overturning two previous versions of the E.U.-U.S. privacy agreement. In both cases, the Court of Justice of the European Union — the bloc’s highest court — agreed with him that the deals failed to protect Europeans’ data from U.S. intelligence gathering.
Schrems urged the E.U.’s executive body in a letter on Tuesday to “immediately take all necessary steps to allow European citizens and businesses an orderly exit” from the agreement. But if not, Schrems wrote, “we also want to inform you, that we are also planning to challenge” the deal “before the courts.”
Markus Lammert, a spokesman for the European Commission, said on Tuesday that the commission would “carefully analyze any implications” the Supreme Court ruling might have for “the EU-U.S. agenda.”
Schrems is optimistic that he won’t need to take the legal route. “Today there was already some momentum building in the E.U. institutions, so we could see some movement,” he told DealBook.
But if not, Schrems said, he plans to file his challenge this summer.
U.S. tech giants are already facing uncertainty across the Atlantic. The European Commission recently proposed a “tech sovereignty” law that would effectively shut U.S. cloud providers out of the bidding for many large public contracts in Europe.
Together AI lands an $8 billion valuation
Two recent trends are reshaping the artificial intelligence boom: Chinese open-source models are closing the gap with American models, and more U.S. companies are turning away from “tokenmaxxing” to rein in soaring costs.
That created an opportunity for Together AI, a San Francisco start-up that provides businesses with a platform to gain access to hundreds of open-source models while improving those models’ efficiency.
On Wednesday, Together AI announced an $800 million funding round at an $8.3 billion valuation, Niko Gallogly is first to report.
The round was led by Prosperity7 Ventures, the venture arm of Saudi Aramco, Saudi Arabia’s state oil company. Nvidia, Vista Equity Partners, General Catalyst and others also participated.
Companies’ “margins are being eaten by token costs on closed-models,” like those offered by OpenAI and Anthropic, Vipul Ved Prakash, Together AI’s co-founder and C.E.O., told DealBook.
As more companies adopt open-source models, which can be used and modified freely, Together AI’s business has surged. In the past year, the start-up has seen a 10,000-fold increase in the number of tokens it processes per month.
The numbers behind the open-source trend:
-
Chinese open-source models can cost one-fiftieth as much as their American counterparts on a per-token basis, according to a recent JPMorgan Chase report.
-
Open-source processing on the A.I. model marketplace OpenRouter rose to 65 percent in June, from 34 percent in January, according to a Citi analysis that Reuters reported.
Some experts worry about U.S. companies’ adoption of Chinese open-source models, citing concerns that the models are built by companies connected to China’s government and that the developers have illicitly leveraged American technology.
But for many customers, the models’ lower cost is enticing.
Market movers
Global stocks are under pressure on Wednesday after investors racked up huge second-quarter gains that point to bumper trading profits for Wall Street.
The S&P 500’s best quarter in six years was aided by a staggering run in chip stocks, especially memory chipmakers, as traders found new darlings in the artificial intelligence rally. But bears see a fragile U.S.-Iran cease-fire, high interest rates and, yes, A.I. bubble worries as potentially spoiling the party in the second half of the year.
The big winners were the PHLX semiconductor index and Korea’s tech-heavy KOSPI index. Both are packed with chipmakers, including Nvidia, Micron, SK Hynix and Samsung, who have thrived amid a spending spree by top A.I. companies, known as “hyperscalers.”
The flip side: Negotiations on a framework peace deal in Iran have sent oil prices plunging, pushing Brent crude, the international benchmark, briefly below prewar levels last week.
Brent tumbled 38 percent last quarter, according to Deutsche Bank data.
Crypto is also in a tailspin. Bitcoin fell 15 percent last quarter, its third straight quarterly decline, the bank said, as retail investors look for better returns elsewhere, including in A.I. stocks.
But crypto has helped fuel President Trump’s family fortunes. The president’s most recent mandatory financial filing shows that he netted a staggering $2.2 billion in revenues last year — $1.4 billion of which came from his family’s crypto businesses.
Investors who snatched up $TRUMP memecoins and digital tokens issued by World Liberty Financial, the Trump family’s crypto venture, contributed to that windfall. One investor was the United Arab Emirates, which took a big stake in World Liberty.
A reminder: Some traders lost big in the Trump crypto frenzy.
What’s next? Kevin Warsh, the Fed chairman, is set to speak on Wednesday at the European Central Bank’s annual conference in Sintra, Portugal. Investors will be looking for clues on his outlook for inflation and rates ahead of Thursday’s big jobs report.
Economists forecast another hot set of hiring figures, which could add to calls for the central bank to raise rates soon.
THE SPEED READ
Deals
Politics, policy and regulation
-
Melat Kiros, a democratic socialist, defeated Representative Diana DeGette, a 15-term incumbent, in Colorado’s Democratic primary, as progressive candidates continued to pressure moderate Democrats. (NYT)
-
TikTok is reportedly finalizing a deal to settle another lawsuit claiming that its social media platform is addictive to minors. (Bloomberg)
Best of the rest
We’d like your feedback! Please email thoughts and suggestions to dealbook@nytimes.com.
