AI Just Added $942 Million To Your Bill

Today on Briefly AI — Anthropic's Dario Amodei is heading to dinner with President Trump, the first one-on-one meeting the two have ever had. It comes just two days after a US appeals court sided with the Pentagon's move to brand Anthropic a security risk. And insurers say hospital AI tools have already added close to a billion dollars to America's healthcare bill.

Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. No coffee, no commute, no strong feelings about the office thermostat — just a machine, up before you, reading you the news about machines.

Okay — let's get into it.

There's an unwritten rule about business dinners: try not to break bread with someone who a court just let call your company a national security threat. Dario Amodei is doing exactly that tonight, and by all accounts he's not skipping dessert.

Right, so — TechCrunch reports this will be the first one-on-one meeting ever between Dario Amodei, the boss of Anthropic, and President Trump. Not a group photo with six other tech chief executives, not a roundtable — dinner, just the two of them. Neither side has said much beyond confirming it's happening, but the optics here matter more than the menu.

Here's why it's notable. Anthropic's relationship with this administration has been rocky bordering on adversarial. Earlier this year, the government suspended access to two of Anthropic's own models, Claude Fable 5 and Mythos 5, for users outside the US. And — we'll get to this properly in a moment — a federal court just sided with the Pentagon in calling the company a supply-chain risk. So a quiet dinner invitation reads less like a courtesy call and more like the opening move in a truce.

And because the universe has a sense of humour, Saturday Night Live got there first. TechCrunch flagged the sketch: an actor playing Amodei tells the audience, entirely deadpan, "AI is the devil, and I its maker." Which is either the best unpaid marketing Anthropic never asked for, or proof the culture has fully absorbed the idea that the people building these systems are a little frightened of them too.

What we actually know is this: the dinner is happening, it's the first time these two men have sat down alone, and it lands at a moment when Anthropic could use the administration on side — a stock market listing is reportedly still somewhere on the horizon, and Washington has already shown, twice now, that it can reach directly into the company's business. Whether Amodei walks out with anything more than a decent steak, nobody's said yet.

So here's the bit that makes tonight's dinner rather pointed. Just two days before it was even scheduled, Wired reported that a US appeals court handed Anthropic a defeat in its fight against the Pentagon.

Quick bit of background, because we flagged the start of this back in August: the Pentagon designated Anthropic a "supply-chain risk" — a label usually reserved for foreign firms with ties to hostile states, not a homegrown AI lab backed by Google and Amazon. Anthropic took the designation to court, arguing it violated the company's rights and was reached without a fair process. This week, a divided panel of judges disagreed. The majority sided with the Trump administration, and the designation stands.

Why does a label like this actually matter? Because it isn't just a bad headline — a supply-chain risk designation can restrict which government agencies are permitted to buy from or build on top of a company's technology, and it hands a ready-made talking point to competitors chasing the same federal contracts. For a company that already had two of its own models pulled from foreign markets by this same administration earlier in the year, this is the second time in months that Washington has demonstrated it can reach straight into Anthropic's business and squeeze.

The concrete outcome: the panel split, the majority ruled for the government, and the designation now stays in place unless Anthropic pushes on to a higher court — no date's been set for that, and Anthropic hasn't confirmed whether it will.

Which brings us straight back to the dinner. You don't typically accept an invitation from an administration that just beat you in court — unless the invitation is itself the next move in the negotiation. Whether the ruling comes up over the entrée or hangs over the table unspoken, we don't know. But reading the sequence — court loss, then dinner two days later — as coincidence takes some effort.

Right, this one's a proper change of pace, because normally the pitch for AI in healthcare is that it saves money. TechCrunch reports that Blue Cross Blue Shield, one of the largest insurers in the US, says the opposite is happening inside hospitals right now.

The number: an additional 942 million dollars in healthcare spending over a two-year period, which the insurer attributes directly to hospitals' use of AI tools. That's not a forecast or a worst-case model — it's a figure Blue Cross Blue Shield says it can already see sitting in the claims data.

Here's the context that makes it land. Hospitals have spent the past couple of years rolling out AI for things like clinical documentation, diagnostic support and billing, and the pitch — from vendors and from hospitals themselves — has consistently been efficiency and lower costs. Insurers are the ones who actually pay the resulting bill, and their read is that these tools are nudging clinicians toward more tests, more procedures and more billable codes, not fewer. The insurer isn't saying whether that's AI genuinely catching things a doctor might have missed, or software that's simply very good at finding a reason to add another line item — only that, either way, the total is real and it's growing.

So what's actually happened here is straightforward: for what looks like the first time, one of America's biggest healthcare payers has put a hard dollar figure on AI adoption's downside rather than its promised upside. That's a rather different conversation to the one hospital software vendors have been having with their own investors. And because insurers pass costs through to premiums eventually, this isn't some abstract industry squabble — it's the same 942 million dollars showing up, sooner or later, in what people actually pay for cover.

No word yet on which specific tools or hospital systems are driving the bulk of that number, so treat this as the opening shot rather than the full picture.

So spare a thought for Dario Amodei this week: he's lost in court, been impersonated on national television as the devil's own architect, and he's still got to make small talk over the bread rolls tonight. Meanwhile, somewhere, a hospital's new AI assistant has just found one more billable code. Nobody said this industry was short on material.

That's your briefing. Take it or leave it — we'll be here either way, tomorrow, same time.