The Morning Download: The Last Pre-AI War

Good morning. The attack on Iran is either the last pre-AI war or the first AI war. Either way, it’s not the defining feature of this conflict.

Anthropic’s Claude was put to use by the U.S. as it struck Iran, the Journal reported

Commands around the world, including U.S. Central Command in the Middle East, use Anthropic’s Claude AI tool, people familiar with the matter confirmed. Centcom declined to comment about specific systems being used in its ongoing operation against Iran.

The command uses the tool for intelligence assessments, target identification and simulating battle scenarios even as tension between the company and Pentagon ratcheted up, the people said.

But the Trump Administration’s sweeping dispute with the company, focusing on the extent to which a private entity can limit the way the federal government puts its technology to use, appears to be mostly about the extent to which AI will be used on the battlefield in future wars. Anthropic’s concerns about the potential use of AI for the purpose of mass surveillance are probably more about its near-term capabilities, though.

The role of AI in this war is still developing, and given the breadth and depth of the governance questions that lay before the public, that’s a relief. The public needs more time to think them through, even as the conflict raises their profile and creates a greater sense of urgency to do so. 

Trump Administration shuns Anthropic, embraces OpenAI. On Friday the federal government said it would stop working with Anthropic and designate it a supply-chain risk. OpenAI the same day announced a deal with the Defense Department to have its models used in classified settings — a status until recently held only by Anthropic.

The fallout came after Anthropic refused to let the military use its models for mass domestic surveillance or autonomous weapons. "The United States will never allow a radical-left, woke company to dictate how our great military fights and wins wars," Trump posted on Truth Social.

What next? For Anthropic, losing classified status cuts off a major revenue stream. For OpenAI, it's a win shadowed by questions about guardrails. For government suppliers, it could mean that they show they don’t use Claude at all.

And for the federal government? Here's the Journal:

The company’s Claude models are being used across the government to do everything from summarizing documents to analyzing data. Pausing those functions to swap in alternatives could set back the government’s AI and modernization efforts, AI experts have said.

More on the negotiations. A source familiar with last week’s negotiations tells the Atlantic that the two parties were close to a deal Friday before Anthropic learned that the Pentagon wanted to use its AI to analyze bulk data collected from Americans, a red line for the AI startup.

Big Tech staffers push back. OpenAI Chief Executive Sam Altman said the company’s deal with the Defense Department includes those same prohibitions on mass surveillance and autonomous weapons, as well as technical safeguards to make sure the models behave as they should.

Silicon Valley standoff. But workers at Amazon, Google and Microsoft signed a letter urging their own executives to refuse Pentagon contracts that would enable autonomous weapons or mass domestic surveillance, FT reports. Signatories include the 700,000-member Communications Workers of America and several major tech worker groups.

Banned by Trump, downloaded by everyone else. The Claude app shot to No. 1 on Apple's App Store the day after the Trump administration moved to block its government use, CNBC reports.

Government agencies raise alarm about use of Elon Musk’s Grok chatbot. The Pentagon’s other, other major AI model decision last week in putting Elon Musk’s xAI at the center of some of the nation’s most sensitive and secretive operations has attracted its own fair share of static.  The WSJ reports that officials at multiple federal agencies for months have raised concerns about xAI’s safety and reliability.

A January GSA report found that Grok-4 did not meet the safety and alignment standards required for general federal use. Two weeks ago, the Pentagon's chief of responsible AI resigned, citing concerns that safety had become an afterthought in the rush to expand AI capabilities.

As more companies trumpet AI-driven job cuts, what can survivors expect? The FT talked to current Amazon employees in the wake of layoff rounds that began in October. For many, the picture isn't pretty: Slumping morale, increased workloads, a higher number of  tech incidents and pressure to adopt AI tools that often aren't ready for prime time. Some AWS engineers are being asked to absorb eliminated roles, including technical writing, with AI supposed to fill the gap.

“You’re being asked to achieve the same goals with a third of the people [in the team],” said one senior AWS employee.

For leaders, the AI-driven job cuts offer a new challenges

Managing burnout. A recent UC Berkeley study finds that even as workers offload tasks to AI, they're logging longer hours. Engineers report "AI fatigue," driven by a constant fear of missing the next breakthrough, Bloomberg reports. 

And maintaining trust. Arianna Huffington, CEO of the behavior-change company Thrive Global, tells the Journal that for bosses to maintain trust, they must be transparent with workers about what is happening.

With AI, “it’s not like we know exactly the impact it’s going to have on jobs,” Huffington said in an interview. “But, of course, it’s going to have an impact on jobs.”

Nvidia has a new inference chip. Guess who’s the first customer. Nvidia is unveiling a new inference-focused processor at its GTC conference in San Jose this month, built in partnership with chip startup Groq. OpenAI is set to be a launch customer, using it to power its Codex coding tool, the WSJ reports. The move marks a significant win for Nvidia as the AI market shifts from model training, where its flagship GPUs are dominant to inference, a field that included rivals like Google, Amazon, and Cerebras.

There is such a thing as too much electricity. The data center buildout is threatening grids at both ends: straining electricity supplies on peak demand days, and flooding grids in the event of a malfunction, the WSJ reports. The culprit for the latter is the automatic backup systems data centers use to protect operations. When they detect a grid disturbance, they instantly sever their connection, creating a sudden, destabilizing drop in demand.

On two separate occasions, months apart, dozens of data centers abruptly dropped off the power grid in Virginia, sending grid operators scrambling to avoid damaging power plants.

But we're still getting used to 5G. Nvidia said it’s partnering with Nokia, SoftBank and T-Mobile to build AI-powered 6G networks, a necessary upgrade the chip maker said for the AI era and its many use cases, Bloomberg reports.

The Mideast conflict widend as Israeli military struck Hezbollah targets in Lebanon, responding to an attack from the Iranian-backed group and widening the conflict in the Middle East. (WSJ)

Futures for Brent crude oil, the global energy benchmark, rose over 7%, fueled by fears of a protracted closure of the key Strait of Hormuz corridor, and attacks on Mideast energy infrastructure. Stock futures pointed to declines of 1% or more for the Dow industrials, S&P 500 and Nasdaq-100. (WSJ)

On Sunday alone, President Trump and his allies offered at least two separate objectives for the assault on Iran, muddying Washington’s intentions for ending a conflict that has engulfed the Middle East and killed three American service members. (WSJ)

Multiple women alleged to prosecutors that they were sexually assaulted by high-profile associates of Jeffrey Epstein, according to the recently released Epstein files. None of the men were charged. (WSJ)

The WSJ CIO Journal Team is Steven Rosenbush, Isabelle Bousquette and Belle Lin.

The editor, Tom Loftus, can be reached at thomas.loftus@wsj.com.