use

President Trump on Friday directed federal agencies to stop using technology from San Francisco artificial intelligence company Anthropic, escalating a high-profile clash between the AI startup and the Pentagon over safety.

In a Friday post on the social media site Truth Social, Trump described the company as “radical left” and “woke.”

“We don’t need it, we don’t want it, and will not do business with them again!” Trump said.

The president’s harsh words mark a major escalation in the ongoing battle between some in the Trump administration and several technology companies over the use of artificial intelligence in defense tech.

Anthropic has been sparring with the Pentagon, which had threatened to end its $200-million contract with the company on Friday if it didn’t loosen restrictions on its AI model so it could be used for more military purposes. Anthropic had been asking for more guarantees that its tech wouldn’t be used for surveillance of Americans or autonomous weapons.

The tussle could hobble Anthropic’s business with the government. The Trump administration said the company was added to a sweeping national security blacklist, ordering federal agencies to immediately discontinue use of its products and barring any government contractors from maintaining ties with it.

Defense Secretary Pete Hegseth, who met with Anthropic’s Chief Executive Dario Amodei this week, criticized the tech company after Trump’s Truth Social post.

“Anthropic delivered a master class in arrogance and betrayal as well as a textbook case of how not to do business with the United States Government or the Pentagon,” he wrote Friday on social media site X.

Anthropic didn’t immediately respond to a request for comment.

Anthropic announced a two-year agreement with the Department of Defense in July to “prototype frontier AI capabilities that advance U.S. national security.”

The company has an AI chatbot called Claude, but it also built a custom AI system for U.S. national security customers.

On Thursday, Amodei signaled the company wouldn’t cave to the Department of Defense’s demands to loosen safety restrictions on its AI models.

The government has emphasized in negotiations that it wants to use Anthropic’s technology only for legal purposes, and the safeguards Anthropic wants are already covered by the law.

Still, Amodei was worried about Washington’s commitment.

“We have never raised objections to particular military operations nor attempted to limit use of our technology in an ad hoc manner,” he said in a blog post. “However, in a narrow set of cases, we believe AI can undermine, rather than defend, democratic values.”

Tech workers have backed Anthropic’s stance.

Unions and worker groups representing 700,000 employees at Amazon, Google and Microsoft said this week in a joint statement that they’re urging their employers to reject these demands as well if they have additional contracts with the Pentagon.

“Our employers are already complicit in providing their technologies to power mass atrocities and war crimes; capitulating to the Pentagon’s intimidation will only further implicate our labor in violence and repression,” the statement said.

Anthropic’s standoff with the U.S. government could benefit its competitors, such as Elon Musk’s xAI or OpenAI.

Sam Altman, chief executive of OpenAI, the company behind ChatGPT and one of Anthropic’s biggest competitors, told CNBC in an interview that he trusts Anthropic.

“I think they really do care about safety, and I’ve been happy that they’ve been supporting our war fighters,” he said. “I’m not sure where this is going to go.”

Anthropic has distinguished itself from its rivals by touting its concern about AI safety.

The company, valued at roughly $380 billion, is legally required to balance making money with advancing the company’s public benefit of “responsible development and maintenance of advanced AI for the long-term benefit of humanity.”

Developers, businesses, government agencies and other organizations use Anthropic’s tools. Its chatbot can generate code, write text and perform other tasks. Anthropic also offers an AI assistant for consumers and makes money from paid subscriptions as well as contracts. Unlike OpenAI, which is testing ads in ChatGPT, Anthropic has pledged not to show ads in its chatbot Claude.

The company has roughly 2,000 employees and has revenue equivalent to about $14 billion a year.

WASHINGTON — New whistleblower documents detail substantial cuts by the Trump administration to the training requirements for new immigration officers.

Among the cuts are the elimination of practical exams, use of force and legal training courses, and an overall reduction in training time, contrary to an official’s testimony to Congress earlier this month.

The documents, provided to Sen. Richard Blumenthal (D-Conn.) by whistleblowers from the Department of Homeland Security, were publicly revealed ahead of a forum Monday afternoon with congressional Democrats — the third in recent weeks probing what the members view as abusive and illegal tactics used by federal agents.

Lauren Bis, deputy assistant public affairs secretary at DHS, said no training hours have been cut.

“Our officers receive extensive firearm training, are taught de-escalation tactics, and receive 4th and 5th Amendment comprehensive instruction,” she said. “The training does not stop after graduation from the academy. Recruits are put on a rigorous on-the-job training program that is tracked and monitored.”

Blumenthal’s office also disclosed the identity of one whistleblower: Ryan Schwank, an attorney who most recently served as an instructor for new Immigration and Customs Enforcement recruits at the ICE Academy within the Federal Law Enforcement Training Center in Georgia. Schwank, who resigned Feb. 13, is scheduled to testify at the forum.

Schwank is one of two whistleblowers who made a confidential disclosure to Blumenthal’s office last month regarding an ICE policy allowing agents to enter people’s homes without a judicial warrant.

In excerpted quotes from Schwank’s prepared testimony shared with The Times, he calls the training program “deficient, defective and broken.”

“Deficient training can and will get people killed,” he wrote. “It can and will lead to unlawful arrests, violations of constitutional rights, and a fundamental loss of public trust in law enforcement. ICE is lying to Congress and the American people about the steps it is taking to ensure its 10,000 new officers faithfully uphold the Constitution and can perform their jobs.”

Blumenthal’s office did not confirm whether Schwank or the other, still anonymous whistleblower provided the documents released Monday in a 90-page memorandum from minority staff of the Senate Permanent Subcommittee on Investigations.

The documents show ICE has eliminated more than a dozen practical exams that ICE officers previously needed to graduate. In July 2021, a cadet needed to pass 25 practical exams to graduate. Now, nine are required.

Eliminated exams include “Judgment pistol shooting,” “Criminal encounters,” and “Determine removability.”

“All of these are now instead evaluated, if at all, mainly by open-book, multiple-choice written exams and without any graded practical examinations,” the memo states.

Comparisons between the program’s syllabus table of contents and general information sections from July 2025 — before the surge in hiring — and this month show that ICE appears to have cut whole courses, such as use of force simulation training, U.S. government structure, criminal vs. removal proceedings, and use of force.

Earlier this month, acting ICE Director Todd Lyons testified to Congress that while the agency had reduced the number of training days to 42 from 75, “We went from five days a week to six days a week. Five days a week was five eight-hour days. We’ve gone to six 12-hour days.”

But the documents appear to contradict Lyons’ testimony.

“The schedules reflected on these documents indicate that current ICE recruits receive nearly 250 fewer hours of training than previous cohorts of recruits,” the memo states.

The training reductions come as ICE plans to bring up more than 4,000 new Enforcement and Removal Operations officers this fiscal year, which ends in September. One of the documents notes that ICE had graduated 803 new officers in 2026 as of Jan. 29 and projected 3,204 more graduates by the end of the fiscal year.

In a written statement, Blumenthal encouraged more whistleblowers to come forward.

“We know about the Trump Administration’s decimation of training for immigration officers and its secret policy to shred your Constitutional rights because of the brave Americans who are speaking out today,” he wrote. “They are coming to Congress because we have the responsibility to not only bear witness to these crimes, but to do something to make sure they don’t happen again.”

WASHINGTON — The Environmental Protection Agency on Friday reapproved the weed killer dicamba for use on genetically modified soybeans and cotton, a pesticide that has raised widespread concern over its tendency to drift and destroy nearby crops.

The agency said dicamba was critical for farmers who would otherwise have their crops threatened by fast growing weeds. To ensure the pesticide is used safely, the agency said it imposed strong protections and limits on its use.

Dicamba is a common weed killer that can be sprayed on top of genetically engineered crops. It kills the weeds but doesn’t hurt the crops. It has been in use for decades, but it has become more widespread on farms in recent years.

Advocates sharply criticized the agency, saying they are moving forward after courts blocked similar efforts in 2020 and 2024. Allowing its use on these two common crops will drastically expand how much is applied and increase harm, advocates say.

Kelly Ryerson, an activist with the Make American Healthy Again movement that has forged a fragile political allegiance with the Trump administration, said she was disheartened by the decision.

“A top priority of mine was to have the use of Dicamba for over-the-top applications permanently discontinued because” of their harm, she said. “New restrictions on use are not sufficient, and will perpetuate the chemical treadmill where many farmers are trapped.”

The EPA said growers want the weed killer and they need to be supported — and that it isn’t a MAHA versus EPA issue.

The agency said concerns about dicamba drifting to places where it was not intended are real and must be managed. It set limits on how much can be applied per acre, how much can be applied on hot days and established buffer zones to prevent harm to nearby crops. If followed, the chemical can be used without threatening humans or the environment, according to EPA.

The American Soybean Assn. applauded the decision, saying clear rules would help farmers prepare for the next growing season and control destructive weeds.

Environmental groups said dicamba drift has damaged immense acreage, devastating vegetable farms, trees and other critical plants.

“When push comes to shove, this administration is willing to bend over backward to appease the pesticide industry, regardless of the consequences to public health or the environment,” said Nathan Donley, environmental health science director at the nonprofit Center for Biological Diversity.

Environmentalists said the EPA’s use restrictions are insufficient, allowing application for too much of the time and for too many days of the year. The buffer the agency uses to prevent harm to nearby plants has already proved ineffective, they said.

Researchers have been working to better understand its health risks. A 2020 study in the International Journal of Epidemiology found that dicamba exposure was linked to an increased risk for some cancers, including liver cancer and a type of leukemia affecting the blood and bone marrow.

Bayer, a manufacturer of dicamba, said the federal registration will allow them to now seek state approvals. They’ll launch training for applicators in the coming weeks.

Phillis writes for the Associated Press.