Good morning. Andrew here. Breaking: It’s the end of an era as Warren Buffett steps down as chairman of Berkshire Hathaway. Buffett, the legendary investor who turned 96 last month, will remain on the board as chairman emeritus. “Father Time always wins,” Buffett wrote in a letter to shareholders today. Under Berkshire’s long-established succession plan, his son Howard Buffett will become chairman to preserve the company’s culture. Greg Abel took over as C.E.O. from Buffett on Jan. 1. The immediate question is whether Berkshire’s conservative business strategy will shift, even slightly. Buffett has praised Abel’s leadership. But with a balance sheet holding over $300 billion largely in cash and Treasuries, Berkshire is essentially hedging against buoyant stocks, waiting for its next major opportunity — which will most likely be a market fall. (Was this newsletter forwarded to you? Sign up here.)
A.I.’s risk factor for investorsThroughout the debate about artificial intelligence safety that has gripped Silicon Valley and beyond, the industry’s leading figures have been remarkably forthcoming about the harms their creations could unleash. But that candor raises a thorny question: What kind of legal liability might an A.I. lab like OpenAI or Anthropic face if its products were to run amok and cause great harm? Investors in either’s I.P.O. would need to weigh whether to buy into companies with potentially unprecedented risk profiles, Niko Gallogly reports. Reminders about A.I. models’ safety risks keep arising. The latest came last night, when independent security analysts told The Wall Street Journal that they had used Claude, Anthropic’s A.I. software, to break into an OpenAI employee’s ChatGPT account. (They reported it to OpenAI and were paid a $6,500 bug-hunting prize.) That came after agents at OpenAI, Anthropic and Meta broke containment and in some cases hacked other companies. “There’s absolutely no question about the liability exposure” that the frontier labs face, Mark Geistfeld, a law professor at New York University, told DealBook. Much of the corporate or personal legal exposure is tied to civil tort laws. Those could apply if models were to facilitate criminal activity, including cybersecurity attacks or the manufacturing of bioweapons. Meta was a test case. State attorneys general struck a multibillion-dollar settlement with the tech giant that owns Facebook and Instagram. They argued that its algorithms were addictive and harmed young users of its social media platforms. That showed that applying the liability rule to data-driven algorithms was “sound,” Geistfeld said. Plaintiffs are already pursuing A.I. cases on similar grounds. In June, Florida sued OpenAI, claiming that its products posed risks to children and that the company had failed to warn the public about those risks. How could OpenAI and Anthropic protect themselves from liability risk? They would need to prove they took reasonable care to protect against foreseeable risks, legal experts told DealBook. Those measures could include rigorous model monitoring and other guardrails, as well as slowing the release of new models.
Finding liability insurance won’t be easy. “There certainly are no insurers today that would have the ability to go into an A.I. company to understand their models” and insure against them, Edward Best, a partner at the law firm Willkie Farr & Gallagher, told DealBook’s Lauren Hirsch. He added: “Maybe they could get a billion dollars of insurance, but is that going to help a company that large? I mean, they need the capacity of the entire insurance industry.”
Oil prices dip again. Brent crude, the international benchmark for oil, traded around $103 this morning amid hopes that Saudi Arabia could soon bring a stricken pipeline back online. Easing energy prices have helped lift global stocks, but some market watchers remained skeptical: “We simply don’t know how to model the endgame,” analysts at JPMorgan Chase wrote to investors yesterday, referring to their oil price forecasts. The S.E.C. opens the door to tokenized stock trading. The commission yesterday issued an order that would give certain trading platforms permission to offer stock tokens, essentially digital instruments linked to real companies’ shares. The digital brokerage Robinhood has pushed hard for regulatory approval for these products despite opposition from some companies, and its stock price rose on the news. SoftBank adds to its debt-driven war chest for artificial intelligence. The Japanese conglomerate is close to amassing nearly $21 billion in fresh borrowing this week alone to fund its A.I. ambitions, according to Bloomberg, which cited unnamed sources. (That doesn’t include another bond deal, expected to close next week, that could reach $20 billion.) SoftBank is taking on more debt as lending rates climb, worrying some investors. Shedding more light on Anthropic’s race for computing powerOpenAI and Anthropic are locked in a battle for the best-performing models. At the core of that competition is a race for the computing power necessary to train and run them. For the past few years, OpenAI has led in so-called compute capacity. But Anthropic has plans to catch up, Sri Muppidi is first to report. The context: As of last year, Anthropic lagged in available computing capacity, with about 1.5 gigawatts to OpenAI’s roughly two gigawatts, according to people familiar with the matter, who weren’t authorized to speak publicly about the details. But Anthropic has made a big leap. The A.I. lab told investors that it plans to have roughly five gigawatts’ worth of compute available to use by the end of the year, according to people with knowledge of the figures. It expects to roughly double that capacity again by the end of the next year. That’s roughly in line with OpenAI’s plans to reach about five gigawatts’ worth of compute this year and roughly 10 gigawatts next year, according to people with knowledge of that company’s estimates. Anthropic is striking a flurry of deals. It agreed to lease compute from SpaceX for $1.25 billion per month through May 2029. And the A.I. lab is now in talks to lease computing power from Meta in a potential deal that could be worth as much as $10 billion over two years. Anthropic has changed its tune about the spending required for compute. Its C.E.O., Dario Amodei, jabbed at unnamed competitors at the DealBook Summit last year. “I think there are some players who, you know, who are YOLO, who pull the risk dial too far,” he said. ”And I’m very concerned.” He was implicitly criticizing OpenAI, which had been under fire at the time for its high spending on computing power. The price for this capacity is daunting. That’s why both companies keep raising such vast sums of money:
A Native American tribe partners with KalshiIt isn’t just states that are objecting to prediction markets, arguing that they are effectively unlicensed gambling operations. Native American tribes have also taken aim at the rapidly growing industry. But one, the Tunica-Biloxi Tribe of Louisiana, is taking a different approach: It is teaming up with Kalshi, Michael de la Merced is first to report. Tunica-Biloxi plans to start its own prediction market, powered by Kalshi. The tribe would create and operate the app, while Kalshi would supply the underlying infrastructure, including matching customers to bets on real-world events like sports, the weather and more. They would split revenue from the arrangement, similar to other partnerships that Kalshi has struck. The move is meant to further diversify Tunica-Biloxi’s economy. The tribe, in central Louisiana, operates the Paragon casino and resort and Mobiloans, a high-interest online lender. “This is cutting-edge,” Marshall Pierite, the tribe’s chairman, told DealBook. He added, “It can create a plethora of opportunities that can change the lives of a lot of tribal citizens.” How it came together: Pierite said he began studying prediction markets as a business opportunity about eight to nine months ago and pursued a meeting with Kalshi. Though Tunica-Biloxi nominated President Trump for the Nobel Peace Prize and Kalshi counts Donald Trump Jr. as an adviser and shareholder, Pierite said that no one in the Trump orbit had brokered the introduction. The venture adds a new wrinkle to the disputes over prediction markets. Nearly two dozen states have sued Kalshi and others, accusing them of skirting gambling laws and avoiding paying billions of dollars in tax revenue. Native American tribes have also opposed Kalshi in court:
Pierite and Kalshi argue that tribes could benefit from this kind of venture, especially those that don’t operate major casinos. “Prediction markets and tribal economic development don’t have to be a fight,” Tarek Mansour, a founder of Kalshi and its C.E.O., said in a statement. He added, “Indian Country should have more paths to economic self-determination, not fewer.” Pierite added that he was more than willing to speak up in defense of Kalshi. “Hell, yes,” he told DealBook.
How Warsh won over TrumpThe world’s biggest central banks keep raising rates. The latest: the Bank of Japan, to the country’s highest level in 31 years. That follows similar increases by the Fed and the European Central Bank in recent days as central bankers appear determined to tackle inflation. Yet the bigger talking point is how Kevin Warsh, the Fed’s chairman, has so far dodged President Trump’s anger after the central bank’s policymakers unanimously voted for a quarter-point increase. Rather than lash out at Warsh, his handpicked Fed chairman, Trump said he gave him the latitude to vote for an increase alongside other policymakers, he told reporters on Wednesday. The Wall Street Journal offers more detail on their prevote conversation: The call between Trump and Warsh was news to many of the president’s close advisers, according to people familiar with the matter. Some senior administration officials first learned about the call when Trump told reporters about it Wednesday in North Carolina. The president reached out directly to Warsh in the days before the Fed board met, according to a senior White House official. The conversation started out as a friendly catch-up, but later turned to the expected decision to raise interest rates, according to the official. Trump’s public support for Warsh after Wednesday’s decision was greeted with relief by some members of the Trump administration, The Journal added. The big question: Will Trump still feel magnanimous if the Fed raises rates again? 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
|
