Business

1. AI guardrails coming
2. Meta’s $1B man leaves
3. Europe hikes, Fed to follow?
4. Scott Kirby wants a deal
5. Fox-Roku under scrutiny

Existential AI fears broke containment this week after an Anthropic researcher quit, warning that frontier labs are “gambling with our lives.” One of his colleagues chimed in, putting the odds of human extinction from AI at more than 10%. Sheryl Crow entered the conversation.

We at Semafor Business aren’t tech experts, but we do know IPOs, and “we might cause the end of civilization” is a doozy of a risk factor. I’m edge-of-my-seat curious how the lawyers at Wilson Sonsini, working on the Anthropic listing, and Cooley and Wachtell, for OpenAI’s, will handle this section of the offering documents. Rohan and I took a crack:

Risk Factors

1. There is no guarantee that the company’s products will not precipitate the collapse of civilization. Such an outcome, which may not require any intentional misuse or manipulation of our products, is, in the view of the Company’s management, an acceptable cost of pursuing its noble mission. Also, China.
2. It is not possible, based on the information available to the Company at the time of this issuance, to determine whether this is a Good or Bad Idea. The Company, under the oversight of the Calamity Calibration Committee of its Board of Directors, continues to assess the likelihood of the end of the world.
3. Key man risk: We have one and he is not popular. Including this key man, none of our employees understand how we create our products or how they work (See Risk Factor 3). Many of those closest to any semblance of comprehension may quit to go live in underground bunkers.
4. At the time of this issuance, we do not carry an investment-grade credit rating, which may increase our borrowing costs or force us to borrow someone else’s and hope nobody notices. We did try to strong-arm the ratings agencies into granting an exception in the name of generational progress, but they are bizarrely holding out.

AI safety bills are coming

A bipartisan AI safety bill may be introduced as early as next week, as Senate staff solicits feedback from frontier labs and advocacy groups, Semafor’s Ashley Gold scoops. Details are sparse, but it will face a competing, likely stricter bill from Sen. Bernie Sanders that would ban superintelligence and pause AI development.

OpenAI’s head of policy wrote a blog post in support of national safety rules, though critics accused it of whitewashing its opposition to previous efforts at regulation. OpenAI lent its support to two California bills — narrowly focused on accrediting AI watchdog groups — just hours before Gov. Gavin Newsom signed them.

Top AI researcher leaves Meta

Meta AI researcher Andrew Tulloch, whose reported $1 billion-plus pay package made him an immediate Silicon Valley legend, is leaving the company, Semafor’s Reed Albergotti scooped yesterday.

Tulloch had been delaying his departure until the company released its new AI assistant, Muse, this week. It is unclear why he left or whether he plans to continue his AI research at another firm. With the AI race in full gear, the top thinkers are in high demand; the question is how long that lasts before the talent pool expands — top universities are churning out AI majors — and paydays adjust.

Every central bank for itself

The European Central Bank raised interest rates. Will the Fed follow? A quarter-point increase in baseline borrowing costs across the bloc came as officials warned of “upside” inflation risks and oil prices shot above $100 a barrel. Fed-funds traders are putting 64% odds that the Fed will do the same next week (arbitrage alert: the chances on Kalshi are a bit lower). But the ECB’s only job is to manage inflation, while the Fed has a trickier dual task of managing both prices and employment, and so is keeping an eye on a labor market stuck in a “low hire, low fire” defensive equilibrium.

Global central banks would prefer to move their interest rates together to avoid warping investment flows and currency markets. (Back in 1986, a Fed compromise was struck to let Germany and Japan cut first.) The post-Covid responses to inflation and, now, slower growth have upset that lockstep instinct as central banks tend to their own economies.

Scott Kirby still wants a big deal

United Airlines CEO Scott Kirby really wants to buy another airline. Just one problem: “We can’t get it done without a willing partner, and we don’t have one,” he told Semafor’s The CEO Signal show. It’s been five months since Kirby penned an open letter to American Airlines laying out his rationale for a deal, a bit of theater we noted was becoming a trend.

American was uninterested, but Kirby is undeterred: He didn’t deny interest in a merger with Delta, as The Wall Street Journal reported, or JetBlue, though he downplayed the latter’s needle-moving potential: “If you’re going to go through that brain damage [of M&A], my first choice is do something big.”

It’s been more than a decade since a wave of mergers collapsed eight US airlines into four. A Government Accountability Office report found higher prices following those deals, but low-cost carriers’ continued woes — Spirit is in bankruptcy and Frontier has lost money five of the past six years — argue in favor of scale.

“Air travel is not a commodity,” Kirby said. “I am 100% certain that [a United-American] merger would be good for customers.”

DOJ probes Fox’s $22B Roku deal

Federal regulators have launched a deeper probe into Fox’s planned $22 billion acquisition of Roku, days after the company parted ways with star anchor Maria Bartiromo, Semafor scooped. Fox confirmed the request for additional data and documents in a regulatory filing.

Opponents of the deal are concerned about the impact it would have on free, ad-supported streaming apps, and about the insights it would give Fox into users’ watching habits; 100 million households have a Roku device. The firing of Bartiromo prompted well-wishes from President Donald Trump and consternation from Trump adviser Peter Navarro, who wrote that the White House had expressed displeasure to Fox News CEO Suzanne Scott.

Elsewhere in antitrust news: DOJ is investigating whether Nvidia tried to sidestep antitrust rules by striking a licensing deal with AI startup Groq, rather than an outright takeover, The New York Times reported.

➚ BUY: Hooks. AI music startup Suno released new models tied to Warner Music and BMG catalogs. Sony and Universal are still fighting Suno in court, but Warner’s CEO defended the collaboration: “If [users] can’t do it here, they’re going to move on to something else,” he told Semafor’s Mixed Signals show last month.

➘ SELL: Covers. SpaceX’s stock price spiked ahead of new shares being released from the post-IPO lockup, suggesting short-sellers are scrambling to cover their bets.

Companies & Deals

• Eastern promise: Abu Dhabi sovereign wealth fund Mubadala took a stake in China’s Luckin Coffee. Mubadala has been increasing its investments in China, though nearly half of its portfolio is still in the US.
• Reality check: Workplace collaboration software Miro, once valued at $17.5 billion, is being sold to Bending Spoons for less than one-tenth of that. It follows Bending Spoons’ bargain takeover of Airtable last month.

Watchdogs

• Gear shift: Ford CEO Jim Farley disputed Transportation Secretary Sean Duffy’s comments that the auto giant was overly reliant on China. “We’re not enabling the Chinese to come here,” Farley told WSJ.

Markets

• Burn notice: Traders are testing Treasury Secretary Scott Bessent’s warnings not to bet against the house: The 10-year Treasury yield hit a three-year high this week despite Bessent’s plan to buy more Treasury bonds to keep borrowing costs down.
• Neither literally nor seriously: But the bond market shrugged at Trump’s proposed $5,000-per-household dividend, which would be bigger than the pandemic-era checks that sparked runaway inflation.