Andrew here. The politics of A.I. just got far more complicated. A new Chinese large language model appears to match, or perhaps exceed, some of the abilities of leading U.S. offerings.
This could change the calculus for American lawmakers. Until now, the risk of falling behind was hypothetical. Now, regulators have a tangible competitor to factor in. More below.
A catalyst for more M.&A.?
Comcast became a cable and media giant over years of M.&A. across two continents. It has just announced its latest deal — a breakup of itself.
But the proposed split raises the question of whether the two parts of Comcast will become buyers, or sellers, Michael de la Merced writes.
What’s happening: Comcast said on Monday that it planned to spin off NBCUniversal and Sky, its European media business, as a separate publicly traded company.
What remains of Comcast will be its broadband operations, which reach more than 65 million U.S. homes and businesses, and global tech platforms.
The revolving-door details:
-
Mike Cavanagh, Comcast’s co-C.E.O., will lead NBCUniversal when it’s spun off.
-
Michael Angelakis, a former Comcast C.F.O. who left the company in 2015, will return to become Comcast’s C.E.O.
-
Brian Roberts, who engineered many of the acquisitions that made Comcast what it is today, will continue to be involved in both companies.
Comcast’s stock has been a laggard for years:
-
Cord-cutting has eroded the long-term viability of cable TV. That prompted Comcast to spin off several networks into Versant at the beginning of the year.
-
More important, its broadband division has lost an enormous number of customers — over 700,000 last year — as wireless and fiber rivals are growing. Comcast has been responding by bundling wireless services with cable.
-
And Peacock, NBCUniversal’s streaming service, is considered a minnow compared with Netflix. (That was a major reason Comcast pursued Warner Bros. Discovery.)
-
Meanwhile, NBC is paying more and more to buy and keep sports rights.
Comcast’s stock is down 56 percent over the past five years, though Monday’s news sent its shares up over 20 percent in premarket trading.
What could NBCUniversal do next? It may seek to buy up smaller assets, like Lionsgate, Starz or AMC Global Media.
But what if someone sought to buy a newly independent NBC Universal? Could Netflix, which also lost out in the Warner Bros. Discovery race, try again? (Its investors may be wary.) Or could Amazon, Apple or Disney? (That could raise antitrust questions.)
And would Comcast finally pursue a deal with Charter Communications? It would combine the nation’s biggest cable players, who have no overlap, and give them the scale to better compete with wireless and fiber rivals. The cost savings could be enormous.
But it would take time, given that Charter is planning to buy Cox Communications for $34.5 billion. And competition officials may still look askance at combining the nation’s two big broadband operators.
And there’s the political angle: President Trump is no fan of Roberts, who owns a big stake in the parent company of MS Now. Would his administration seek to stymie any major move by a Roberts-affiliated company?
HERE’S WHAT’S HAPPENING
U.S. and Iranian officials will hold talks on Tuesday in Qatar, President Trump says. His announcement was the latest indication that tensions were de-escalating after the two sides traded strikes over the weekend, including over a standoff in the Strait of Hormuz. Iran’s claim of sole control over the waterway, along with mines placed in it, threaten to derail a return to normal shipping volumes through the vital energy passageway. Oil prices edged higher as traders assessed the state of an uneasy cease-fire.
A dangerous heat wave is expected to hit parts of the central and eastern U.S. Forecasters said that temperatures could break records in some areas, like New York, Philadelphia and Washington. The extreme weather come as Europe is experiencing struggling with a record-shattering heat wave, with officials in France reporting that the heat had already been responsible for roughly 1,000 excess deaths.
All eyes will be on the Supreme Court. The court is slated to release the final rulings of its term this week, including decisions on Trump’s bid to fire the Fed governor Lisa Cook and his effort to restrict the guarantee of birthright citizenship. Markets will also focus on the jobs report for June, to be released on Thursday, for signs of a heating labor market.
South Korea will build a giant chip-making center. The South Korean government, Samsung and SK Hynix announced that they would invest nearly $600 billion in a new complex that will include four semiconductor factories. The prices for memory chips are expected to rise up to 50 percent over the next three months, potentially forcing more consumer electronics companies to follow Apple and increase prices.
Chinese A.I. is catching up
The Trump administration has let Anthropic rerelease its powerful Mythos model but just to a handful of government-approved partners. And OpenAI must release its latest model only in a restricted preview.
The moves underscore how an administration that had been largely hands-off toward the artificial intelligence industry is doing a 180. And that interventionist streak is coming as critics worry about Chinese competition.
What’s new: Security researchers have found that GLM-5.2, an open-source model by the Chinese developer Z.ai, has caught up to Mythos when it comes to finding cybersecurity bugs, according to The Wall Street Journal. Some caveats: It lags Mythos in other abilities, and Anthropic’s less-powerful Opus 4.8 can accomplish some of the same feats.
Those advances come as Washington keeps limits on advanced U.S. models. Mythos remains accessible to only select customers. And Fable, a related model that is meant to have more safety guardrails, remains unavailable.
OpenAI said that its latest model, GPT-5.6 Sol, was also restricted to a small number of companies at first. “This isn’t quite the process that we think is optimal,” Sam Altman, the C.E.O. of OpenAI, wrote on X of the conditions. He added that OpenAI was working with the government and that it “shares most of our goals.”
The danger for U.S. companies: a more interventionist Washington may put them at a disadvantage to Chinese rivals who tend to make open-source and freely available models. Western companies — including Microsoft and Coinbase — are at the least pondering adopting Chinese models given their balance of high power and low cost.
Consider: Over the past week, six of the top 10 most-used models on OpenRouter, a popular A.I. model exchange, were Chinese. (GLM-5.2 ranked 7th.)
What may be next: Anthropic and others are reportedly urging Congress to crack down on Chinese open-source models that, they say, are being illicitly trained on their own.
Salaries for law grads jump higher
Big law firms are enjoying staggering upticks in profitability after years of consolidation and double-digit billing rate increases. That has led to eye-popping compensation for top partners despite concerns that artificial intelligence tools could disrupt the industry.
Now the elite firms are again ratcheting up the pay scale for entry-level lawyers in a competitive market, Elizabeth Olson reports for DealBook.
Raising the bar: Milbank moved first to raise associate salaries earlier this month when it said it would increase the base salary for first-year lawyers by $10,000 to $235,000 annually. It will now pay lawyers in their eighth year — usually when associates make partner or leave — a base of $455,000.
Sullivan & Cromwell has reportedly decided to match Milbank’s salary scale, according to Above the Law, with first-year associates at $235,000.
The compensation growth is a top-down phenomenon. The consolidation trend means firms have been divvying profits between fewer owner-partners. And leading firms have had a bigger financial pot to shell out for rainmakers, the lawyers who attract big-name business.
That’s led “king-of-the-hill firms to pay top partners like major league sports figures, with salaries crossing the $30 million and $40 million threshold for a handful,” Kent Zimmermann, a legal industry consultant at Zeughauser Group, told DealBook.
Turnover among firms has also accelerated. Equity partners once stayed around longer to form the backbone of firm reputation. But in 2025, a record 83 percent of associates who left their firms departed in the first five years. Firms have also taken on more non-owner partners, who are paid less and have less security.
To shore up their foundations, big firms are aggressively recruiting law students as early as their first year in law school, according to a survey of 3,000 law students by the National Association for Law Placement, a legal professional service, and the Law School Admission Council.
“Large law firms compete fiercely for top talent, which is not only pushing the recruiting timeline earlier, but also increasing the cost of acquiring talent,” Nikia Gray, the N.A.L.P.’s executive director, told DealBook.
Quote of the Day
“Doooonnnk, doooonnnk, doooonnnk.”
A trimmed and slowed-down audio clip of a commentator at an e-sports event in Bucharest, Romania, posted by someone seeking to show that he had said “Donk,” the nickname of a star player. More than $5.7 million had been wagered on Polymarket on whether the word would be used during the event.
Human managers cut A.I. employees more slack
Corporate America’s embrace of artificial intelligence has begun to give the technology a human sheen, with many companies building A.I. bots into org charts and calling them as “employees.”
But recent research, including from Emma Wiles, a professor who studies A.I.’s effect on workers, shows that corporate managers aren’t good at being objective about these new workers. They appear to be biased against work by humans, for example, preferring work done by A.I.
And that bias may be paving the way for a big flaw in how managers oversee A.I. agents, Noam Scheiber reports for The Times:
In an experiment involving dozens of companies with A.I. employees, the researchers found that managers tended to vet documents less carefully when told an A.I. employee had produced them. The managers missed errors that other managers caught when told they were vetting the work of a human.
Dr. Wiles speculated that managers didn’t think sussing out mistakes made by A.I. employees was their responsibility. If something went wrong, they could dismiss it as the fault of the tech team, or of the executives who wanted A.I. employees in the first place. “But it’s not your problem,” she said, channeling the managers’ mind-set about their own roles.
The bigger problem, Scheiber writes, is that most corporate users don’t know about their own biases. “Several scholars said the impact of these undetected biases could grow,” he writes. “One way is if future models are trained on data produced by today’s models without sufficient care, creating a kind of self-reinforcing loop.”
THE SPEED READ
Deals
Politics, policy and regulation
Best of the rest
We’d like your feedback! Please email thoughts and suggestions to dealbook@nytimes.com.
