Latest / Connor with Honor: AI, Real Estate and Real Life / The Pentagon Says Anthropic Could Sabotage Military AI. Where Is the Evidence?

Transcript

The Pentagon Says Anthropic Could Sabotage Military AI. Where Is the Evidence? · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: Yesterday, the U.S.
• 0:00
  SPEAKER_00: government filed a 40-page legal brief in federal court.
• 0:04
  SPEAKER_00: The argument, and I want you to hear this clearly, was that Anthropic might secretly disable or alter its own AI.
• 0:11
  SPEAKER_00: That's during active warf war fighting operations, if they disagreed with how it was being used.
• 0:17
  SPEAKER_00: That's the U.S.
• 0:17
  SPEAKER_00: Department of War, arguing that a San Francisco AI company is a potential saboteur of American military operations in a court filing yesterday.
• 0:27
  SPEAKER_00: This thing is getting wilder by the day.
• 0:30
  SPEAKER_00: This is AI with honor.
• 0:31
  SPEAKER_00: Today's Wednesday, March 19th.
• 0:33
  SPEAKER_00: Let's get it done.
• 0:34
  SPEAKER_00: Here's where we are.
• 0:35
  SPEAKER_00: Anthropic sued the U.S.
• 0:36
  SPEAKER_00: government after being designated a supply chain risk, a label that's never been applied to an American company before.
• 0:43
  SPEAKER_00: Anthropic went to court and asked for a preliminary injunction, meaning stop enforcing this designation while we fight it out.
• 0:51
  SPEAKER_00: The Department of War filed its rebuttal yesterday.
• 0:55
  SPEAKER_00: 40 pages.
• 0:56
  SPEAKER_00: Their core argument, and I want to be precise here because this is important, is that Anthropic might attempt to disable its technology or alter the behavior of its model before or during wartime operations if the company feels its corporate red lines are being crossed.
• 1:12
  SPEAKER_00: Let that land for a moment.
• 1:14
  SPEAKER_00: U.S.
• 1:14
  SPEAKER_00: government is arguing in federal court that the maker of Claude, the AI used to generate business proposals, write code, and answer customer questions, and run voice agents for mortgage companies, might secretly sabotage American military operations mid-battle.
• 1:30
  SPEAKER_00: So no evidence was actually cited to support that concern.
• 1:33
  SPEAKER_00: The lawyer representing Anthropic told TechCrunch there had been no investigation, no incident, nothing.
• 1:39
  SPEAKER_00: But here's what's happening underneath that argument.
• 1:42
  SPEAKER_00: The government is terrified of a world where a private company has a kill switch on technology embedded in classified military systems.
• 1:50
  SPEAKER_00: And honestly, that's a legitimate fear.
• 1:53
  SPEAKER_00: The problem is the remedy that they chose.
• 1:56
  SPEAKER_00: Designating a U.S.
• 1:58
  SPEAKER_00: company as a foreign adversarial level threat, that's the part that's without precedent.
• 2:04
  SPEAKER_00: The hearing on Anthropic's request for a preliminary injunction is set for March 24th, next Tuesday.
• 2:11
  SPEAKER_00: So let's watch for that date.
• 2:12
  SPEAKER_00: And one more thing, multiple tech companies, employees at OpenAI, Google, and Microsoft, along with legal rights groups, have led Amisk briefs in support of Anthropic.
• 2:22
  SPEAKER_00: The industry is not letting this go quietly.
• 2:24
  SPEAKER_00: Anthropic revenues just tripled.
• 2:27
  SPEAKER_00: And while they're fighting the government in court, the business is exploding.
• 2:31
  SPEAKER_00: Anthropic's annualized revenue recently closed at 19 billion and crossed actually 19 billion.
• 2:37
  SPEAKER_00: And here's the number that should stop you cold.
• 2:40
  SPEAKER_00: That's nearly triple what it was at the end of last year.
• 2:43
  SPEAKER_00: Up 36% in just the PAX two weeks alone off of this controversy.
• 2:49
  SPEAKER_00: One source told Fortune that if you extrapolate from just the most recent week of revenue, the run rate is closer to$30 billion annualized.
• 2:58
  SPEAKER_00: Think of what this means.
• 2:59
  SPEAKER_00: The government tried to hurt them, and they actually turned Anthropic into one of the fastest growing companies in the history of technology.
• 3:07
  SPEAKER_00: Can a guy get on lot B, please?
• 3:09
  SPEAKER_00: OpenAI just gave free users a serious upgrade.
• 3:13
  SPEAKER_00: So OpenAI rolled out GPT 5.4 mini and nano to chat GPT free users this week.
• 3:19
  SPEAKER_00: These are smaller, faster, more cost-efficient versions of the flagship model.
• 3:24
  SPEAKER_00: The move is calculated.
• 3:25
  SPEAKER_00: More users on the platform means more data, more data means better models, and puts competitive pressure on every other AI company trying to hold a paid subscriber base.
• 3:34
  SPEAKER_00: This is also the clearest sign of what happens when AI becomes a commodity.
• 3:38
  SPEAKER_00: The best tool six months ago is now the free version.
• 3:41
  SPEAKER_00: Jensen Wong just told us that the next five years what that's going to look like.
• 3:45
  SPEAKER_00: GTC wrap Monday.
• 3:47
  SPEAKER_00: And now that the dust is settled, here's what actually matters from the keynote.
• 3:50
  SPEAKER_00: Jensen Wong walked on the stage and said NVIDIA expects purchase orders for its Blackwell and Vera Rubin chip platforms to reach$1 trillion, trillion, trillion dollars through 2027.
• 4:01
  SPEAKER_00: Last year that projection was$500 billion.
• 4:03
  SPEAKER_00: He doubled it on stage in front of 30,000 people.
• 4:06
  SPEAKER_00: The new Vera Rubin platform pairs an NVIDIA CPU with the Rubin GPU and a tightly integrated architecture built specifically for AI inference.
• 4:15
  SPEAKER_00: Not training, but inference.
• 4:17
  SPEAKER_00: And that's the shift.
• 4:18
  SPEAKER_00: The industry spent years building systems to train the AI models.
• 4:22
  SPEAKER_00: Now the money is moving to running them fast, cheap at scale, and constantly.
• 4:28
  SPEAKER_00: The Groke 3 LPU just dropped.
• 4:30
  SPEAKER_00: Nvidia paid$20 billion to acquire Groke's technology last year.
• 4:35
  SPEAKER_00: This is the first chip out of the deal.
• 4:38
  SPEAKER_00: It's built to accelerate inference specifically.
• 4:40
  SPEAKER_00: One rack holds 256 of these chips and sits beside the Vera Rubin rack as a combined inference powerhouse.
• 4:48
  SPEAKER_00: And then Jensen talked about Feynman, the architecture after Rumen, coming in 2028, built on a 1.6 nanometer process.
• 4:56
  SPEAKER_00: For context, the current smallest process is 3 nanometers.
• 5:00
  SPEAKER_00: This is doubling of density.
• 5:02
  SPEAKER_00: More power, less heat, faster, everything.
• 5:05
  SPEAKER_00: Jensen also validated OpenClaw, the open source AI agent platform that went from one developer side project to the most starred project on GitHub in weeks.
• 5:15
  SPEAKER_00: He also called it a seismic shift in how humans will interact with AI going forward.
• 5:20
  SPEAKER_00: The through line of the entire keynote was this: NVIDIA just doesn't make chips.
• 5:25
  SPEAKER_00: It wants its own operating layer of the entire agentic AI economy.
• 5:30
  SPEAKER_00: Chips, software, infrastructure, inference, physical AI, robotics, all of it.
• 5:34
  SPEAKER_00: One stack, one company in the center of every layer.
• 5:38
  SPEAKER_00: That's either visionary or terrifying, depending on where you're sitting.
• 5:42
  SPEAKER_00: Anthropic just published the largest AI user study ever.
• 5:46
  SPEAKER_00: Yesterday, the same government, the same day government filed a brief against them.
• 5:50
  SPEAKER_00: Anthropic published the results of a study they ran with Claude users.
• 5:54
  SPEAKER_00: 81,000 people participated, the largest and most multilingual qualitative study of AI ever conducted.
• 6:00
  SPEAKER_00: They asked three questions.
• 6:02
  SPEAKER_00: How do you use it?
• 6:03
  SPEAKER_00: How what do you dream it could make possible?
• 6:05
  SPEAKER_00: And what do you fear it might do?
• 6:07
  SPEAKER_00: The full results are at anthropic.com.
• 6:09
  SPEAKER_00: I'll be digging into those numbers for an upcoming episode, but the timing of that publication, the same day the government called them a national security threat, was not accidental.
• 6:18
  SPEAKER_00: That's the company saying, here's what our users actually are.
• 6:22
  SPEAKER_00: Here's who they are, here's what they actually want, here's what they think, and here's the trust we've actually built.
• 6:29
  SPEAKER_00: So let's get into some token cost.
• 6:31
  SPEAKER_00: Here's where AI pricing is right now and what it actually means for your life and your business.
• 6:36
  SPEAKER_00: We watched a video announced a trillion dollars projected chip sales.
• 6:40
  SPEAKER_00: That's OpenAI giving away their best model from six months ago for free.
• 6:44
  SPEAKER_00: Anthropics revenue tripled in a year.
• 6:46
  SPEAKER_00: Jensen Huang is already talking about the chip, the next chip generation after this one.
• 6:52
  SPEAKER_00: Most people watching this are trying to figure out whether AI is actually useful for anything real, and here's the honest answer.
• 6:58
  SPEAKER_00: It's like the year in the internet, 1997.
• 7:00
  SPEAKER_00: If you told a small business owner in Santa Clarita that they needed a website, most of them would have said they don't see the point.
• 7:06
  SPEAKER_00: The phone works fine, the Yellow Pages ad was pulling leads, the referral network was solid.
• 7:11
  SPEAKER_00: By 2005, the businesses without websites were losing ground, they couldn't explain and couldn't get back.
• 7:18
  SPEAKER_00: That's exactly where we are with AI right now.
• 7:20
  SPEAKER_00: Except the timeline is much more compressed.
• 7:23
  SPEAKER_00: We're not talking about an eight-year window, we're talking about 18 to 24 months before the gap between operators who built AI and operators who didn't become structural, meaning you can't just close it by trying harder.
• 7:35
  SPEAKER_00: When I sit down with real estate agent or mortgage companies or brokerage who want to talk about AI, the first thing I ask them isn't what tools you've looked at.
• 7:42
  SPEAKER_00: It's what does your follow-up system look like, 11 p.m.
• 7:45
  SPEAKER_00: on a Saturday when a lead comes from your website?
• 7:48
  SPEAKER_00: Because that's where the money's leaking.
• 7:50
  SPEAKER_00: It's not the fancy stuff.
• 7:51
  SPEAKER_00: It's the gap between where a potential client raises their hand and the human being responds.
• 7:56
  SPEAKER_00: An AI agent fills that gap every time for every lead at a cost that's trivial compared to with what a missed transaction would end up costing you.
• 8:06
  SPEAKER_00: The good news, the tools are better than they've ever been, cheaper than they've ever been.
• 8:10
  SPEAKER_00: What NVIDIA just announced means inference costs, the cost of actually running AI responses will keep dropping as Ruben and Grok 3 come online.
• 8:19
  SPEAKER_00: It's like Moore's Law applied to intelligence, instead of just transistors.
• 8:23
  SPEAKER_00: The bad news is cheap and accessible means your competition has access to the same tools.
• 8:28
  SPEAKER_00: The differentiator isn't the tool, it's the architecture.
• 8:31
  SPEAKER_00: Just knowing what problem to solve first.
• 8:33
  SPEAKER_00: It's building a system that compounds instead of one that just adds noise to your workflow.
• 8:37
  SPEAKER_00: And the ugly part of it, OpenAI gave their powerful model away for free to everybody with a phone.
• 8:43
  SPEAKER_00: That includes the 25-year-old agent in your market who has nothing to lose all day than to experiment.
• 8:48
  SPEAKER_00: The experience gap you spent 20 years building doesn't protect you from someone with better systems.
• 8:54
  SPEAKER_00: It only protects you if you have, well, better systems too.
• 8:58
  SPEAKER_00: And that's a conversation I have every day.
• 9:00
  SPEAKER_00: You can check out more.
• 9:01
  SPEAKER_00: Codedbyconnor.com.
• 9:03
  SPEAKER_00: Anyway, come have it.
• 9:04
  SPEAKER_00: Let's get into uh AI agentic in business as well.
• 9:07
  SPEAKER_00: The anthropic situation just changed what enterprise AI looks like.
• 9:11
  SPEAKER_00: Here's something that hasn't been covered enough.
• 9:13
  SPEAKER_00: In the wake of the Pentagon situation, Anthropic expanded its enterprise AI agent push.
• 9:19
  SPEAKER_00: New clawed integrations across Slack, DocuSign, FactSec, and Gmail.
• 9:24
  SPEAKER_00: All in the same week.
• 9:25
  SPEAKER_00: While we were fighting the government in court or while they were, they were simultaneously deeping the route, deepening the side business tools they run on every day.
• 9:33
  SPEAKER_00: This isn't by accident.
• 9:35
  SPEAKER_00: I don't think anything that they do is.
• 9:37
  SPEAKER_00: This is a strategy.
• 9:38
  SPEAKER_00: When one door closes, built in more.
• 9:41
  SPEAKER_00: And software stacks and stocks rebounded in the news.
• 9:45
  SPEAKER_00: The market read it correctly.
• 9:46
  SPEAKER_00: Anthropic losing a government contract doesn't hurt the business.
• 9:50
  SPEAKER_00: Anthropic embedding Claude into Slack and DocuSign, those are integrations that will be compounding over years.
• 9:56
  SPEAKER_00: Think about that from a positioning standpoint.
• 9:59
  SPEAKER_00: The businesses that are winning in this environment aren't the ones chasing the loudest story.
• 10:03
  SPEAKER_00: They're the ones quietly building infrastructure while everybody else watches the drama.
• 10:08
  SPEAKER_00: That's the model.
• 10:09
  SPEAKER_00: That's what iHelp companies do.
• 10:12
  SPEAKER_00: AI is now buying elections.
• 10:14
  SPEAKER_00: This one's definitely worth talking about because the Washington Post reported that AI company money is flooding into the 2026 midterms.
• 10:22
  SPEAKER_00: One of the 20 candidates in recent Texas and North Carolina primaries who received AI industry funding.
• 10:29
  SPEAKER_00: Only one lost.
• 10:31
  SPEAKER_00: AI companies are not building projects, just building products anymore.
• 10:35
  SPEAKER_00: They're actively shaping the political environment they operate in.
• 10:39
  SPEAKER_00: The anthropic Pentagon fight, open AI DOD deal, the legislation being written or not written in Washington, none of this is happening in a vacuum.
• 10:48
  SPEAKER_00: There's our billions of dollars being spent to ensure the regulatory environment stays favorable.
• 10:53
  SPEAKER_00: Understanding that is part of the understanding the landscape you're building in.
• 11:05
  SPEAKER_00: Maybe that was on purpose.
• 11:07
  SPEAKER_00: Jensen Huang confirmed NVIDIA has received order from China and is starting manufacturing for that market.
• 11:13
  SPEAKER_00: The export restrictions that were designed to limit Clauds or China's access to frontier AI chips, NVIDIA's finding a path around them through chips calibrated to stay just below the restriction thresholds.
• 11:24
  SPEAKER_00: This is geographical AI race in real time.
• 11:27
  SPEAKER_00: U.S.
• 11:27
  SPEAKER_00: government restricts exports.
• 11:29
  SPEAKER_00: Nvidia engineers chips that compete technically but deliver meaningful capability.
• 11:34
  SPEAKER_00: China gets access to compute, and the race, of course, continues.
• 11:38
  SPEAKER_00: There's no clean ending to this story.
• 11:40
  SPEAKER_00: Tech companies are pledging to cover AI energy costs.
• 11:43
  SPEAKER_00: The AI center build-out is creating real political pressure around electricity prices.
• 11:48
  SPEAKER_00: Community near data centers are seeing grid strain.
• 11:52
  SPEAKER_00: Voters are noticing in the utility bills.
• 11:54
  SPEAKER_00: In response, OpenAI, Microsoft, Google, Amazon, Meta, and Oracle have all pledged to cover that energy infrastructure costs associated with their AI data center so that the consumers don't see higher energy bills.
• 12:06
  SPEAKER_00: Well, that that's a remarkable moment.
• 12:08
  SPEAKER_00: The largest technology companies in the world voluntarily agreeing to absorb an infrastructure cost that would otherwise hit regular people, specifically to avoid a political backlash heading into the midterms.
• 12:20
  SPEAKER_00: AI is now an energy story, a political story, an election story, and it's still just getting started.
• 12:26
  SPEAKER_00: And why don't they use some of that and also take care of the people that AI is going to displace?
• 12:31
  SPEAKER_00: The preliminary injunction hearing for Anthropic versus the U.S.
• 12:35
  SPEAKER_00: government is set for next Tuesday.
• 12:37
  SPEAKER_00: That's the moment a federal judge decides whether to temporarily block the Pentagon supply chain risk designation while the lawsuit plays out.
• 12:45
  SPEAKER_00: If the injunction is granted, anthropic gets breathing room and the case moves towards trial.
• 12:51
  SPEAKER_00: If it's designed, the designation stays in place while the fight continues.
• 12:56
  SPEAKER_00: Contractors keep being required to certify that they don't use clawed in military work.
• 13:01
  SPEAKER_00: Industry groups representing hundreds of companies are urging the court to pause the pause of the blacklisting.
• 13:07
  SPEAKER_00: The argument that the Pentagon could have simply ended the contract rather than applying a national security designation is one of the multiple legal experts and even former Trump administration officials agree with.
• 13:19
  SPEAKER_00: The tech industry is watching Tuesday like a hawk because the precedence this sets about whether the government can use procurement designations as punishment for companies that don't do what they're told goes far beyond anthropic.
• 13:33
  SPEAKER_00: It goes to every company building anything in or near the government space.
• 13:37
  SPEAKER_00: I want to leave you with something from the Anthropic study that published yesterday.
• 13:41
  SPEAKER_00: Now, 81,000 people, the largest quantitative AI study ever run, and Anthropic asked them, what do you dream AI could make possible?
• 13:49
  SPEAKER_00: And the answers weren't about productivity, they were about time, getting more of it back, about spending and all the on the things that matter.
• 13:57
  SPEAKER_00: People aren't asking for and make them rich, they're asking for AI to give them their life back.
• 14:02
  SPEAKER_00: I think about that in the context of what I do every day.
• 14:05
  SPEAKER_00: I build systems that take the follow-up, the intake, the lead capture, the qualification and scheduling off the plate of people who got into the real estate or mortgage or their profession because they wanted to help people, not because they wanted to manage a CRM at midnight.
• 14:21
  SPEAKER_00: That's the promise AI done right.
• 14:23
  SPEAKER_00: Not replacing the human, but giving the human back, getting the human back to the only they can do.
• 14:29
  SPEAKER_00: The government wants to talk about kill switches and war fighting operations.
• 14:32
  SPEAKER_00: 81,000 regular people just told us that they want to have more time with their family and with themselves.
• 14:39
  SPEAKER_00: That's the real use case.
• 14:41
  SPEAKER_00: That's what we're building towards.
• 14:43
  SPEAKER_00: Connor with Honor.com, I will see you tomorrow.
• 14:46
  SPEAKER_00: Thanks for watching.