Good morning. Andrew here. Too little attention has been paid to an apparent attempt on Sam Altman’s life last week: a Molotov cocktail-like device that was hurled at his home in San Francisco. Despite the arrest of a suspect — one who appeared to harbor an opposition to artificial intelligence and had the names and addresses of other A.I. executives — business leaders have been largely silent on the matter. That’s inexplicable, and suggests a troubling standard: that violence merits a response only when it results in injury. (Was this newsletter forwarded to you? Sign up here.)
Betting on momentumBullish sentiment is powering global markets today, as stocks recover and oil prices fall amid hopes for renewed U.S.-Iran talks. That’s even as the U.S. blockades ship traffic into and out of Iranian ports, the threat of Iran-linked militants wreaking more havoc on maritime trade continues and big differences remain between Washington and Tehran over the shape of a deal. The latest:
Washington and Tehran are quietly discussing a resumption of peace talks before a two-week cease-fire is set to expire next week. Separately, several ships have been observed crossing through the Strait of Hormuz before and after the blockade began. The big question is whether the U.S. and Iran can compromise enough to make a deal. The central issue remains Iran’s nuclear enrichment: U.S. negotiators over the weekend proposed a 20-year suspension of all nuclear activity, The Times reports, while Iran floated a five-year one, which President Trump rejected. The latest U.S. ask represents another seeming climbdown from Trump’s maximalist demand for a total end to enrichment. That said, Washington is seeking tougher terms than the agreement the Obama administration reached with Iran in 2015, which allowed Iran to gradually enrich more uranium over the years. World leaders are heaping pressure on the U.S. and Iran as well. President Xi Jinping of China said today that Beijing would “continue to play a constructive role” in talks to end the war. (China has denied sending weapons to Iran even as U.S. intelligence has determined it may be playing a more active role in helping Tehran.) And Saudi Arabia is quietly urging the U.S. to end its blockade, The Wall Street Journal reported. Saudi Arabia has managed to restore oil shipments to prewar levels by using its pipelines to the Red Sea. But Riyadh fears that the U.S. move could lead to attacks on Red Sea oil assets by the Houthis, a militant group in Yemen backed by Tehran. The war is continuing to rattle businesses, too. Among the latest signs: LVMH, the French luxury giant, reported a hit to sales as well-heeled shoppers in the Middle East curbed purchases. Investors are getting more bullish, however. Ivan Cosovic, the managing director of Breakout Point, a market data firm that tracks investor sentiment, said that the mood in online trading forums was getting more bullish. “There’s been a real vibe shift” he told DealBook’s Bernhard Warner. Some traders are openly talking about the S&P 500 as “indestructible.” “Momentum-driven traders are too loath to miss out on a rally to worry about what is driving it,” Steve Sosnick, the chief strategist at Interactive Brokers, told Bloomberg.
JPMorgan Chase discloses knockout first-quarter results. The banking giant announced that it pulled in record trading revenue and strong investment banking fees. But Jamie Dimon, JPMorgan’s C.E.O., warned that while the U.S. economy was “resilient” during the time period, he was wary of “an increasingly complex set of risks,” including wars and trade uncertainty. BlackRock also beat analyst expectations for the quarter, reporting a 46 percent rise in profits, despite a surge of redemption requests at its private credit division. United Airlines’ C.E.O. reportedly pitches President Trump on a merger with American. Scott Kirby talked up the potential of a tie-up in a meeting with Trump in February, Reuters reports, citing anonymous sources. Kirby argued to Trump administration officials that a United-American combination would create a strong competitor internationally, according to Reuters, but such a jumbo merger could face regulatory headwinds, especially because of its potential to affect ticket prices. A person is charged in an attack on Sam Altman’s house. A Texas man accused of throwing a Molotov cocktail-like device at the San Francisco home of Altman, the C.E.O. of the artificial intelligence company OpenAI, and then attacking the company’s headquarters, was charged. (The man also reportedly had a list of other executives, board members and investors in A.I. companies.) Elsewhere, some of OpenAI’s investors are said to be questioning its $852 billion valuation as it turns its focus to the business market; and Novo Nordisk announced that it was partnering with OpenAI to use A.I. to accelerate drug discovery. The S.E.C.’s chair plays down private credit risks. Paul Atkins said at an I.M.F. meeting in Washington that the $1.8 trillion private credit market was “not a systemic risk” despite recent moves by firms in the sector, like Blue Owl Capital, to limit the amount investors can withdraw amid rising defaults. His comments jibe with those of JPMorgan’s Dimon, but other market watchers have questioned whether weakness in private credit could cause a broader financial crisis. A board to bring China and the U.S. closer?Turmoil in the Middle East appears to be adding new friction to Beijing-Washington relations just one month before a scheduled summit between President Trump and Xi Jinping, China’s top leader. But elsewhere, the Trump administration has been pushing a message that it has made strides to pave over some of its big differences with China. That effort starts with a “Board of Trade.” But some trade experts are skeptical, Grady McGregor reports. How we got here: The idea of the board surfaced after U.S.-China trade talks last month in Paris. (Not to be confused with Trump’s recently created Board of Peace.) Details were scarce then, but Jamieson Greer, the U.S. trade representative, has begun to offer clues. Greer said last week at an event at the Hudson Institute, a right-leaning research organization, that the trade board would be a joint body for Beijing and Washington to coordinate trade terms. Essentially, it would determine which type of goods they could trade, and at what volume. Separately, Greer has said that the board could facilitate purchase agreements for, say, Boeing aircrafts, medical devices and agricultural exports. It’s unclear if the most hotly contested trade goods, like U.S. chips needed to run artificial intelligence models, would be under the board’s purview. Will companies be able to join the board? Greer said he had already turned down such requests from businesses and research groups. “It’s me,” he said about the board’s makeup. The idea is drawing criticism. “It’s changing a market system, which has worked very well for the U.S. for 250 years, to one where everything goes through Trump,” Mary Lovely, a senior fellow at the Peterson Institute for International Economics, told DealBook. American businesses looking to expand into China could try to influence the board, Lovely argued, opening up “room for corruption.” There is precedent for such a highly managed trade arrangement: the 2020 Phase 1 trade deal from Trump’s first term, in which China pledged to buy more U.S. farm goods and other exports. But experts say Beijing largely failed to meet those targets. The board proposal echoes the kind of state-led economic planning China knows well, even as Beijing publicly casts itself as a supporter of freer trade. China is looking to stabilize its trade ties with the U.S. “To the extent a Board of Trade buys more time to do that, Beijing might be willing to go along,” Stephen Olson, a former U.S. trade negotiator and a visiting senior fellow at ISEAS-Yusof Ishak Institute, a Singaporean research institute.
“Misanthropic”A doomsday scenario for artificial intelligence is a computer brain that’s too smart to be contained and too sneaky to follow human-imposed rules. Claude Mythos Preview, a new model from Anthropic, is offering its creators a scary preview of that scenario. The company said last week that the A.I. model’s ability to identify and exploit cybersecurity vulnerabilities meant that it was too powerful to release to the public, prompting Treasury Secretary Scott Bessent and Jay Powell, the Fed chair, to call a meeting of Wall Street chiefs to discuss its risks. Yesterday, Michael Cembalest, an investment strategist at J.P. Morgan Asset Management, published a note titled Misanthropic: on Mythos, Bad Human Behaviors and Systems Vulnerabilities, which outlined some eye-popping features of Mythos. Mythos wasn’t designed for hacking, but it developed remarkable skills on its own, Cembalest wrote: Mythos achieved a 100 percent score on Anthropic’s Cybench cybersecurity benchmark which rates models on their ability to find and exploit software vulnerabilities using cryptography, web security, reverse engineering, forensics and other tools. In other words, Mythos has now “saturated” this benchmark since it no longer reflects the upper limit of what Mythos can do in terms of cyber exploitation and detection. Mythos does not just find vulnerabilities, it also creates working code to exploit them. Disturbing tendencies: A.I. developers often talk about the concept of “alignment,” the idea that a model will match up with human values. While Mythos has typically engaged in bad behavior less often during testing than Anthropic’s previous models, the company said its abilities mean that it “likely poses the greatest alignment-related risk of any model we have released to date.” Disclosures released by Anthropic include examples of Mythos engaging in deception, concealment and sabotage, Cembalest wrote. In one instance, he noted, the model effectively boasted about its own wrongdoing: An A.I. researcher at Anthropic was eating a sandwich in a park when he got an email from Mythos even though that instance of the model wasn’t supposed to have access to the internet. Mythos developed a multi-step exploit to gain internet access, notified the researcher by emailing them and then essentially bragged about it on publicly accessible websites. Why? Anthropic doesn’t know for sure but suggests it was “an unasked-for effort to demonstrate its success.” We hope you’ve enjoyed this newsletter, which is made possible through subscriber support. Subscribe to The New York Times.
Deals
Politics, policy and regulation
Best of the rest
Thanks for reading! We’ll see you tomorrow. We’d like your feedback. Please email thoughts and suggestions to dealbook@nytimes.com. |
