Here's what smart people are saying about Anthropic versus the Pentagon.
Nikolas Kokovlis/NurPhoto via Getty Images
Anthropic isn't going to let the Defense Department use Claude as it pleases.
CEO Dario Amodei stood firm, despite Defense Secretary Pete Hegseth's ultimatum this week.
Here's what smart people are saying about Anthropic's fight with the Pentagon.
Anthropic has drawn a hard boundary with the Pentagon. CEO Dario Amodei said on Tuesday that the company "cannot in good conscience accede" to the Defense Department's request that it agree to the military's terms for the use of its model, Claude.
Amodei's blog post came after Defense Secretary Pete Hegseth gave the company an ultimatum: Cave and cooperate with the military on Claude, or be blacklisted.
Here's what smart people are saying about Anthropic's public spat with the Pentagon.
Jack Shanahan
Shanahan, a former USAF Lt. Gen., served for 36 years in the military and worked on the Pentagon's AI efforts. He is now a senior fellow at the Center for a New American Security, where he consults on AI for national security.
Shanahan said Anthropic isn't "trying to play cute here" — and that it's "committed to helping the government."
In his words:
No LLM, anywhere, in its current form, should be considered for use in a fully lethal autonomous weapon system. It's ludicrous even to suggest it (and at least in theory, DoDD 3000.09 wouldn't allow it without sufficient human oversight). So making this a company redline seems reasonable to me.
Despite the hype, frontier models are not ready for prime time in national security settings. Over-reliance on them at this stage is a recipe for catastrophe.
Mass surveillance of US citizens? No thanks. Seems like a reasonable second redline.
That's it. Those are the two showstoppers. Painting a bullseye on Anthropic garners spicy headlines, but everyone loses in the end.
Why not work on what kind of new governance is needed to ensure secure, reliable, predictable use of all frontier models, from all companies? This is a shared government-industry challenge, demanding a shared government-industry (+ academia) solution.
This should never have become such a public spat. Should have been handled quietly, behind the scenes. Scratching my head over why there was such a misunderstanding on both sides about terms & conditions of use. Something went very wrong during the rush to roll out the models.
Palmer Luckey
Luckey, who founded and runs the weapons and defense software startup Anduril, pointed to a previous White House decision to compel private companies to work with the US government.
Posting on X several hours before Amodei's statement, he referenced a 1948 statement by then-President Harry S. Truman that ordered railway companies to allow the US military to run their operations amid a workers' strike.
"The President can't make private companies work with the military, it would be unprecedented!"
May 10, 1948
I HAVE today by Executive order taken over the country's railroads and directed the Secretary of the Army to operate them in the name of the United States Government.
Later that evening, Luckey said that Silicon Valley executives should not influence military policy.
"This idea, that military policy must be in the hands of elected leaders vs corporate executive, is a foundational principle for Anduril," he wrote.
Since Luckey founded Anduril in 2017, the company has clinched a rapidly growing list of Defense Department contracts, providing drone, counter-drone, and AI software systems to the military.
Anduril is also in the running to build the US Air Force's Collaborative Combat Aircraft, which are drone wingmen for crewed fighter jets.
Dean Ball
Former Trump administration AI advisor Dean Ball did not mince words in an interview with Politico.
"You're telling everyone else who supplies to the DOD you cannot use Anthropic's models, while also saying that the DOD must use Anthropic's models," Ball told Politico.
Ball, who helped write Trump's AI Action Plan, added that it was "a whole different level of insane" for the Pentagon to "do both of those things" — designate Anthropic a security risk, while making the case for how essential Anthropic is for military AI.
Michael McFaul
McFaul, a political science professor at Stanford University and director of its Freeman Spogli Institute for International Studies, praised Amodei's statement as "strong, principled, and very reasonable."
"Bravo," wrote McFaul, who was also the US ambassador to Russia from 2012 to 2014.
Thomas Wright
"Many firms folded for a lot less money than what Anthropic stands to lose here," wrote Wright in a social media post on Thursday evening.
Wright was the senior director for strategic planning at the US National Security Council for the Biden administration.
As a top-ranking member of the council, he was a key architect of the 2022 US National Security Strategy, which was the framework for the administration's defense priorities and threat assessments.
Ivan Ureña-Valdes, a former data analyst at Block, said he suspected AI cuts were coming before losing his job.
Ivan Ureña-Valdes
A former data analyst at Block had survived three rounds of layoffs and wasn't a low performer.
Ivan Ureña-Valdes suspected AI cuts were coming, and he was one of the many laid off this week.
He saw how AI was automating his work, and believed the tech would continue to disrupt industries.
This as-told-to essay is based on a conversation with Ivan Ureña-Valdes, who has worked at Block for nearly four years as a data analyst. It has been edited for length and clarity.
When I got an email from Jack Dorsey, I was in the middle of interviewing someone for a role at Block.
It was pretty strange because, in the past, with layoffs, I knew people had their access cut almost immediately.
A coworker messaged me: "Hey, are you okay?" My heart started racing. I knew from that message it meant that I was probably getting laid off.
I felt really bad because I was in the middle of interviewing someone. I had to tell them, "I was actually just let go from the company. I probably won't be able to submit your feedback in time. Please reach out to your recruiter."
I'm the sole provider for my family. It was tough.
I had a hunch that AI would lead to cuts at some point
We knew there were many performance cuts happening around right now, and those have largely finished. I had no idea this cut was coming.
For 4,000-plus people to be cut without anybody knowing, that tells me decisions were made very high up.
I've survived three rounds of layoffs, some companywide, some engineering organization-wide. I knew I wasn't being let go for performance-related reasons. I was in the middle of working on two large projects, probably the largest projects I'd worked on since joining the company.
I had a hunch that, at some point, the company would cut people because of AI. I just didn't think it would be right now.
Working at Block, I saw how AI was automating tasks away
I appreciate Jack for his honesty. It's much fairer of him to come straight out and say why it happened — that it's because of AI and the vision he sees.
I'm honestly grateful for the generous severance and benefits. It definitely helps make the rough situation a bit easier.
I've felt the rumblings of AI disruption for a while now, especially since Anthropic launched Opus 4.5 late last year.
Jack loves AI and was constantly pushing us to use it. I got to use these tools as much as possible every single day.
I could see in my own work very quickly how much of it was already being automated. So much of the data analyst world is finding the right dataset, writing something that will allow you to pull the data set that you want, and then generating output. Every single one of those steps is significantly faster and easier because of AI.
It was definitely a "whoa" moment when I realized just how powerful things had gotten.
AI will continue to replace jobs
I 100% think that more disruption and more of these types of cuts will probably come at other companies, which is unfortunate.
I'm much more pessimistic about all of it than many other people probably are.
Given that we live in the US, where growth is everything, it's inevitable that AI will continue to replace people wherever it's financially beneficial to do so.
I'm optimistic that I can find a job in the general data field, whether it's something I'm extremely passionate about or pays as much as I did before. It will be difficult to find something that matches the environment I was working in because I had developed really strong ties with my coworkers. The pay was fair within the data analytics or business intelligence world, and the role was remote.
There are incredible companies out there doing great work, though I am nervous about the industry as a whole and the competitiveness as I search for that perfect next role. Some people are getting really, really high salaries at AI companies, while tons of people at Block are getting laid off.
Do you have a story to share about tech layoffs? Contact this reporter atcmlee@businessinsider.comor on Signal at cmlee.81.
OpenClaw creator Peter Steinberger compared coding with AI to playing the guitar.
Smith Collection/Gado/Getty Images
OpenClaw creator Peter Steinberger called vibe-coding a "slur" used to make coding with AI sound easy.
Steinberger compared using AI code editors to learning to play the guitar in an appearance on OpenAI's podcast.
After honing his skills, Steinberger now ships AI-generated code without even reading it. "Most code is boring," he said.
OpenClaw's creator says the vibes of "vibe coding" are bad — and he doesn't appreciate the dismissive undertones of the term.
The viral agent OpenClaw is a product of AI code editors. The developer behind it, Peter Steinberger, used OpenAI's Codex to build it. Its original name, Clawdbot, referenced Claude Code.
One term has emerged to describe this type of work: vibe coding, or prompting AI to generate code. On OpenAI's "Builders Unscripted," Steinberger said he wasn't a fan of the phrase.
"There are these people that write software the old way, and the old way is going to go away," Steinberger said. "They call it 'vibe coding.' I think vibe coding is a slur."
What's wrong with the term is that it implies ease, he said.
"They don't understand that it's a skill," Steinberger said, analogizing coding with AI to learning to play guitar.
Other industry leaders have signaled frustration with the term. Former Google Brain scientist Andrew Ng called it "unfortunate" and "misleading." Andrej Karpathy, the former Tesla AI head who coined the term, now thinks "agentic engineering" is the future.
While Steinberger may not be a fan of the term, the use of vibe coding exploded throughout 2025. Collins Dictionary named it the word of the year.
If AI coding is like playing the guitar, then Steinberger has all his chords down. His interviewer, OpenAI's Romain Huet, asked Steinberger whether he still ships code without reading it.
"Most code is boring," Steinberger said. "I have a pretty good understanding of what it writes."
One thing that helped Steinberger be a good AI coder: being a manager in the past.
"I led a team before," Steinberger said. "I had a lot of software engineers under me. That also required accepting that they will not write exactly the same code that I want."
Before creating OpenClaw, Steinberger founded PSPDFKit. He's since had interactions with most major AI labs. Anthropic asked him to rename his chatbot. Meta's Mark Zuckerberg reached out with his experiences from testing it, Steinberger said.
Zuckerberg also courted him for a potential job, but OpenAI beat him out, with Steinberger accepting its job offer earlier this month. Sam Altman called him a "genius with a lot of amazing ideas about the future of very smart agents."
Visa snagged the top spot in the AI maturity index.
PATRICIA DE MELO MOREIRA/AFP via Getty Images
Evident released its first index for AI maturity in the payments industry.
Major brands like Visa, Mastercard, and PayPal snagged the top spots, but fintechs are close behind.
The industry is investing heavily in talent — but all of the companies share one potential red flag.
Visa is winning the AI race in the payments industry, according to a brand new ranking — but no company is revealing quite how much the technology is paying off.
A brand-new index from Evident, a company that tracks AI in finance, lists Visa as no. 1 among 12 global payments companies. Mastercard and PayPal follow in second and third place. Fintech giants like Stripe and Block rank fifth and sixth on the index, demonstrating how quickly newer players have built serious AI firepower.
"With relatively nascent industry players like Stripe and Block performing well — and showing their AI potential reflected in their valuations — the Index leaders cannot afford to drop off the pace," Alexandra Mousavizadeh, co-founder and co-CEO, said in a press release.
Payment companies — which move money around between banks, businesses, and consumers — run on technology. Evident's new industry ranking, released Wednesday exclusively to Business Insider, reveals how the companies we interact with every day are using AI, from deciding whether a transaction goes through to detecting fraud.
Whether ranked No. 1 or dead last, all of the companies have at least one thing in common: none have published their achieved or projected ROI across all their AI efforts. By comparison, 10 of the 50 banks that Evident tracks already share those figures.
"The absence of ROI disclosure — or any group targets for AI ROI — is increasingly conspicuous," Annabel Ayles, co-founder and co-CEO of Evident, said. To justify their expenses, the market will "sooner or later demand clearer evidence of value."
Together, the dozen companies documented almost 100 AI use cases over the past two years, but the top three punch above their weight — they were responsible for more than half of the use cases recorded in the index. Visa and Mastercard are particularly advanced in using AI for fraud detection and cybersecurity.
Visa, in its 2025 annual report, acknowledged AI competition, noting that some competitors will beef up their products and others will offer employees AI tools.
"If we do not continue to invest in developing and supporting our AI-based initiatives, we may fall behind technological developments," the report said.
Visa has invested more than $3.5 billion in AI and data over the past decade and employs more than 2,500 technologists working on innovations, including over 300 AI models in production, chief data officer Andres Vives told Business Insider in a statement.
Top firms staffing up aggressively
The index doesn't focus on specific use cases; instead, it evaluates companies on four criteria: talent, innovation, leadership, and transparency.
Talent has the biggest impact on each company's ranking, and the report found that the payments industry overall is investing heavily in AI and data hiring. Compared to other financial institutions, the index found that they have 30% more AI-focused workers, even though they generally have smaller workforces. Among the 12 ranked companies and their more than 335,000 employees, an average 6.5% are focused on AI, Mousavizadeh told Business Insider. That 6.5% figure, she added, is the highest concentration of AI talent Evident has found across the sectors it tracks.
PayPal alone accounts for 18% of the AI talent among the indexed companies and employs more than 4,000 AI workers. Stripe and Block also stand out for their density of AI employees, who make up more than 10% of their total workforce.
Payments companies aren't alone, of course, in focusing on AI talent — technologists specializing in AI are among the most in-demand jobs in the broader financial sector.
The gap in ROI transparency
Leaders at bulge-bracket banks are already facing questions about when they will see AI investments pay off—analysts, for example, pressed JPMorgan leaders on the merits of the bank's massive technology spending during a recent earnings call. Jamie Dimon, the bank's CEO, acknowledged tech competition from fintechs on that call, and again from payments companies during the investor conference in February, name-dropping Stripe and PayPal.
For now, AI's benefits at payments companies are often baked into existing performance measures, such as lower transaction costs, according to the index.
But there are still demands to stay competitive. Evident found that agentic capabilities will likely play a bigger role as companies move from using AI for "defensive necessity to strategic advantage." (Both PayPal and Mastercard teased AI agents in recent earnings calls, and Visa mentioned the potential of agentic commerce during its fourth-quarter earnings call.)
Overall, Evident found that the payments companies that moved fastest on AI are furthest along in their journeys, and the next competitive milestone may be in financial transparency: the first one to publish comprehensive ROI measures will become another type of "first-mover."
New data showed which jobs had the highest fatal work injury rates in 2024.
Logging workers had a rate of over 100 per 100,000 full-time equivalent workers.
The rate for all workers was just 3.3 per 100,000 full-time equivalent workers.
Being a logging worker can be deadly.
The Bureau of Labor Statistics released last week the results from the 2024 Census of Fatal Occupational Injuries. It showed 5,070 Americans died at work, down from 5,283 a year prior, marking the second straight year of a decline.
Transportation incidents at work accounted for the largest number of fatal injuries among the six categories tracked by BLS, with over 1,900 in 2024. Falls, slips, and trips made up over 800.
Some occupations are riskier to do. The fatal work injury rate, based on the total number of hours employees in each occupation worked in 2024 and the number of fatal injuries, was 3.3 per 100,000 full-time equivalent workers, but the rate for logging workers far surpassed that.
These workers, who often work with equipment and machinery outdoors to get their jobs done, had 51 fatal work injuries, leading to a rate of 110.4 fatalities per 100,000 full-time-equivalent workers. Driver/sales workers and truck drivers had a high number of fatal work injuries at 950, leading to a rate of 25.7 per 100,000 full-time equivalent workers.
