How the U.S. could (and should) respond to DeepSeek

General Motors CEO Mary Barra told investors yesterday that GM will save up to—Dr. Evil voice—one billion dollars a year by yanking robotaxi funding for unprofitable subsidiary Cruise.

A bitter pill to swallow for Cruise employees, no doubt. But if I were Barra, I wouldn’t save it for a rainy day at the proving ground. With Apple’s CarPlay and Android Auto out of fashion in Detroit, can you just imagine what $1 billion could do for a better infotainment system? —Andrew Nusca

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

How the U.S. will fight back against DeepSeek, according to experts

As DeepSeek’s recently introduced innovations wiped over $1 trillion off the value of U.S. stocks, President Donald Trump said DeepSeek’s rise was “a wake-up call for our industries that we should be laser-focused on competing to win.”

The U.S. government’s immediate reactions shed little light on its plans to ensure U.S. superiority in the AI race. But nobody expects the Trump administration to take its foot off the gas.

“In the near term, DeepSeek’s achievement is likely to pressure the U.S. into increased support for domestic AI development, most likely leading to increased federal investment in AI research and infrastructure,” wrote Raymond James analyst Ed Mills in a Monday note.

Itamar Friedman, CEO of Israeli code-generation startup Qodo, said the Trump administration would do well to invest in smaller companies and academia, now that it appears massive clusters of cutting-edge AI chips are not as necessary for leading research as they seemed just weeks ago.

Trump on Monday threatened to levy new tariffs on chips imported from Taiwan, the center of the global semiconductor industry. At the same time, his administration wants to stimulate U.S. AI development, as exemplified by last week’s announcement of the $500 billion Stargate project.

Those things “are in clear tension,” said Robert Trager, co-director of the Oxford Martin AI Governance Initiative at Oxford University. “If you’re going to put a tariff of 100% on chips coming into the U.S., it’s going to make it much more expensive to build out these data centers [using the latest chips].”

Trager said he expects the White House to maintain Biden administration export controls on the most powerful AI chips from U.S. companies, particularly after DeepSeek admitted that access to chips was its biggest bottleneck.

But it could take another year or two for those export controls to really bite in China, said Sam Bresnick, a research fellow at Georgetown University’s Center for Security and Emerging Technology. So far they have only reduced China’s training capacity by 10% to 15%, he noted, but the capacity gap would naturally grow as more clusters of the latest chips are deployed in the U.S. —David Meyer

Block launches open-source AI agent Goose

Block, the tech company once known as Square, announced yesterday the launch of an open-source AI agent called Goose.

The tool allows developers to customize it for different purposes (from checking code for bugs to sending emails) using different large language models (like OpenAI’s o1 or DeepSeek’s R1). It can be used for any individual or commercial purpose.

Goose also furthers the debate over open- versus closed-source development, brought to the forefront this week by the surprise breakout of DeepSeek. Block and its leader Jack Dorsey, the former Twitter CEO, are champions of the former. 

It’s the “best way to build software,” said Manik Surtani, Block’s open-source lead, in an interview with Fortune.

Goose is the first project out of Block’s new dedicated open-source office, which supports the company’s suite of products, including Cash App, Square, and Tidal. 

Surtani said the technology first came out of engineers building a tool to help them write software, but they realized its potential to perform other types of tasks, like organizing calendars or sending emails.

“Anyone can add to it and extend it,” Surtani said, later adding: “We want open [source] by default to be a North Star.” —Leo Schwartz

RTX now has a $218 billion backlog for aviation, defense

As geopolitical tensions rise around the world, defense contractors like RTX are seeing their sales pick up.

RTX said Tuesday that it now has a $218 billion backlog for its commercial aircraft and defense products—due to an uptick in passenger air travel activity as well as new U.S. and international weapons and defense contracts. 

On a call with investors, RTX CEO Christopher Calio said it had more than 30 defense systems being used in active combat around the world, and that there had been “tremendous” demand for its Patriot missile defense system, F135 propulsion system, and Coyote kinetic effector, among other products.

Part of the backlog—which is 11% larger than it was at the end of 2023—is new business. RTX said it had received more than $112 billion in new contracts in 2024 alone. The defense portion of the backlog grew 19% year-over-year.

Can AI help alleviate the demand? Calio said RTX has been using AI to quicken the pace of its product testing cycles within the company.

Within the company’s avionics department, RTX is using generative AI for things like product testing, first article inspections, and RFP responses, helping cut software testing cycle times within its avionics business by 3x.

RTX plans to use generative AI for another 40 use cases this year, Calio said. —Jessica Mathews

More data

—Another OpenAI researcher quits. Cites ‘very risky gamble’ to pursue AGI.

—Apple chips can be hacked. Two vulnerabilities can leak personal data via web browser.

—AT&T Q4 earnings beat expectations. 1% revenue growth and seven straight years of subscriber growth.

—U.S. Navy tells personnel: Don’t use DeepSeek. Security and ethics concerns, per memo.

—MGM Resorts to pay $45 million to settle 14 lawsuits related to 2019 data breach, 2023 ransomware attack.

—Garmin’s “blue triangle of death.” Activities that use GPS are suddenly bricking smartwatches.

—DeepSeek’s security flaws mean you can instruct it to generate “malicious outputs,” according to one threat intelligence firm.

—Apple to add Starlink support to iOS. T-Mobile is reportedly testing with some users.

—Meta's Reality Labs tops internal targets. 2024 sales up 40%, per internal memo.

Endstop triggered