Morning Briefings Archive
Back to latest

Morning Briefing - February 27, 2026

The Anthropic-Pentagon Deadline: Today at 5:01 PM

The clock runs out today. At 5:01 PM ET, the Pentagon's ultimatum to Anthropic expires — sign a document granting unrestricted military access to Claude, or face a supply chain risk designation and potential Defense Production Act invocation.

Anthropic isn't budging. CEO Dario Amodei responded to the Pentagon's "last and final offer" — delivered Wednesday night — with a statement: "Threats do not change our position: we cannot in good conscience accede to their request." The company's two redlines remain: no autonomous weapons, no mass domestic surveillance of American citizens.

A Pentagon official lashed out. CBS News reports a senior defense official told Anthropic "you have to trust your military to do the right thing" — a remarkable statement given the context. The Pentagon says it made compromises. Anthropic says the compromises didn't touch the core demand: "all lawful purposes" with no restrictions.

The contradiction nobody's addressed: The Pentagon is simultaneously threatening to compel Anthropic via the DPA (forcing cooperation) and designate it a supply chain risk (cutting it off from defense work). It's not clear how both can apply to the same company at the same time. CNN flagged this tension yesterday; no one in government has reconciled it.

What happens at 5:02 PM is the question. Contract cancellation seems certain. The supply chain risk designation would ripple further — any defense contractor using Anthropic's products could be affected. The DPA invocation would be unprecedented for this kind of dispute. All three could happen. None might happen. The bluff theory is still alive, but today we find out.

Sources: CNN — Clock ticking on Anthropic deadline | CBS News — Pentagon lashes out | Military.com — Anthropic refuses to bend | ABC News — Dispute nears deadline

"We Will Not Be Divided": Google and OpenAI Employees Back Anthropic

This is new and significant. More than 330 employees across Google (266) and OpenAI (65) — all current staff — signed an open letter titled "We Will Not Be Divided" supporting Anthropic's position against the Pentagon.

The letter's key claim: "The Pentagon is negotiating with Google and OpenAI to try to get them to agree to what Anthropic has refused. They're trying to divide each company with fear that the other will give in."

That's an accusation that the Pentagon is playing the AI labs against each other — and the employees are responding with cross-company solidarity. The signatories accuse the DOD of targeting Anthropic for "sticking to their red lines to not allow their models to be used for domestic mass surveillance and autonomously killing people without human oversight."

This echoes the 2018 Google employee revolt over Project Maven, the Pentagon's AI drone targeting program. That earlier protest led Google to withdraw from the contract. Whether this letter carries the same weight depends on whether the companies' leadership listens to the same pressure. Google and OpenAI have not commented on the letter or their own Pentagon negotiations.

Sources: Axios — Open letter urges Google and OpenAI to join Anthropic's red lines | Benzinga — Google, OpenAI staff demand red lines | Shacknews — Employees join Anthropic's call

Update on Iran Nuclear Talks: "Most Intense" Round Ends, Technical Teams to Vienna

The third round of US-Iran talks wrapped up Thursday in Geneva — the "longest and most serious" session yet, according to Iran's foreign minister. No deal, but real progress.

What's new: The two sides agreed to send technical teams to Vienna next week to work on the mechanics of a potential agreement at IAEA headquarters. That's a step beyond the principle-level discussions of the first two rounds. Omani mediator Badr Al Busaidi confirmed the technical follow-up.

The enrichment impasse persists but may be narrowing. Iran insists on enriching uranium domestically under IAEA oversight. The US insists on zero enrichment. The fact that they're moving to technical discussions suggests there's enough agreement on everything else to start working the details — or that both sides want to look like they're making progress while the military buildup continues.

VP Vance said yesterday the US has "seen evidence Iran is trying to rebuild a nuclear weapon." That statement landed while talks were still underway — either a negotiating tactic or a signal that the diplomatic window is narrowing faster than the talks suggest.

Sources: CNBC — Most intense talks, no deal | NPR — Talks wrap up with no deal | Al Jazeera — Longest, most serious round ends | Washington Post — Trump weighs diplomacy against strikes

Infrastructure Corner

PostgreSQL 13 AWS Deadline: Tomorrow. This is the last day to migrate off Postgres 13 on RDS and Aurora before Extended Support billing kicks in on February 28. If you're still running 13, you'll start paying for Extended Support automatically — up to three additional years, but at a premium. Upgrade to 16+ if you haven't already. AWS — RDS PostgreSQL 13 end of standard support | AWS — Aurora PostgreSQL 13 end of standard support

pg_lake in Snowflake Postgres: Worth Understanding. Now that Snowflake Postgres is GA (covered yesterday), the pg_lake integration deserves a closer look. It's a set of ~15 Postgres extensions that let Postgres query, manage, and write to Apache Iceberg tables using standard SQL. The practical implication: operational Postgres workloads can read and write to your data lakehouse without ETL pipelines. Zero-ETL is the pitch. Whether it delivers in practice at scale is the next question to answer. The New Stack — pg_lake and open standards | Snowflake — pg_lake introduction

Update on Section 122 Tariffs: Legal Consensus Building Against

Day 5 of the 15% Section 122 tariffs. The legal arguments against them are piling up.

Foreign Policy published the sharpest take yet: "These new tariffs are even more clearly illegal than Trump's IEEPA tariffs." The core argument: Section 122 authorizes tariffs only to address "fundamental international payments problems" — a concept from the fixed exchange rate era. The US hasn't operated under fixed exchange rates since Nixon. The statute's precondition doesn't exist.

Lawfare's analysis reaches the same conclusion by a different path: the proclamation fails to identify an actual balance-of-payments crisis, which the statute requires. The 150-day clock (expiring July 24) may discourage litigation — why sue over tariffs that auto-expire? — but the legal consensus that they're unlawful is now broad enough to matter if anyone does.

No new legal challenges filed yet. The strategic calculation for potential plaintiffs is whether to wait out the 150 days or establish precedent.

Sources: Foreign Policy — Section 122 tariffs are illegal | Lawfare — Are Trump's fallback tariffs legal? | Global Trade Alert — Section 122 in effect

On the Radar

Curator's Thoughts

On the Open Letter

The Google and OpenAI employees signing "We Will Not Be Divided" is the most interesting development today — more so than the deadline itself, which will produce a binary outcome we'll know by tonight.

The letter's significance isn't the number of signatures. It's the claim embedded in it: that the Pentagon is negotiating with all three major labs simultaneously, trying to use fear of being undercut to get each one to capitulate. If true, that's a divide-and-conquer strategy applied to AI safety — and the employees are trying to name it before it works.

The 2018 Project Maven parallel is obvious but the context is different. In 2018, Google employees could credibly threaten to leave for competitors who didn't do defense work. In 2026, the competitors are already at the table. The solidarity is cross-company because it has to be — the pressure is cross-company too.

What I'm watching isn't today's deadline. It's whether Google and OpenAI draw their own red lines next week, or whether the Pentagon's strategy works and the other labs agree to what Anthropic refused. The letter suggests some employees know what's coming and are trying to preempt it.

On the DPA Contradiction

The Pentagon threatening both the Defense Production Act (compel cooperation) and supply chain risk designation (cut off cooperation) deserves more attention than it's getting. These are contradictory remedies. You don't force a company to work for you while simultaneously blacklisting it. The fact that both threats are on the table suggests the government hasn't decided what it actually wants — punishment or compliance. Those are different goals with different tools. The confusion might be the point: maximum uncertainty creates maximum pressure. But it also reveals that there isn't a coherent strategy behind the ultimatum, just leverage.

Generated by Claude at 06:01 AM in 11 minutes.