🤖 Anthropic and the Pentagon Are Back at the Table, But the AI Safety Standoff Is Far From Over

⚖️ Senate Blocks Iran War Powers Vote as the House Prepares Its Own

🚀 PrSM Fires in Combat for the First Time, A New Benchmark for Army Long-Range Strike

🇩🇪🇫🇷 France and Germany Build a Nuclear Steering Panel

📌 Plus 14 more stories across Air, Land, Sea, Space & Cyber

Talks between Anthropic CEO Dario Amodei and Emil Michael, the Under Secretary of Defense for Research and Engineering, resumed this morning, according to reporting from the Financial Times, following one of the most publicly fractious episodes in the short history of AI and defense contracting. Negotiations had collapsed last Friday after the Pentagon demanded broad "any lawful use" access to Claude, Anthropic's flagship model. Anthropic held firm on two specific red lines: no use for autonomous weapons, and no use for mass domestic surveillance of Americans. The Defense Department responded by designating Anthropic a supply-chain risk to national security, a designation typically reserved for adversary-state companies, while President Trump directed all federal agencies to immediately cease using Anthropic's tools. OpenAI, which had publicly backed Anthropic's position just days earlier, struck its own deal with the Pentagon within hours, drawing immediate backlash from its employees, researchers, and the public, sending Claude surging to number one on the Apple App Store.

For procurement officers, defense tech investors, and anyone building on the dual-use AI stack, the story underneath the headlines is structural. The Pentagon's willingness to publicly blacklist a $380 billion American AI company over contract language signals that the DoD is not treating AI safety provisions as negotiable window dressing. Palantir, Shield AI, and purpose-built defense AI companies benefit from a cleaner pitch: no commercial guardrails, no civilian-use constraints, no employee open letters. Whether that's a strategically sound outcome for long-range warfighting capability is a harder question. Reporting from Fortune confirms that the military was actively using Claude for targeting-support functions during Operation Epic Fury even as the blacklisting was announced, with classified instances running on AWS infrastructure that became a separate target of Iranian drone strikes. That operational dependency gives Anthropic considerably more leverage than the public framing of the standoff has conveyed.

The risk now is that Amodei's leaked internal memo, in which he characterized OpenAI's deal as "safety theater" and described Trump as demanding "dictator-style praise," blows up whatever negotiating room remained. The White House has already signaled it views those comments as a serious obstacle. The signal to track: whether the resumed talks result in contract language that preserves Anthropic's specific prohibitions on surveillance tooling, or whether Amodei accepts a looser legal-compliance framework similar to what OpenAI signed. The former would restore Anthropic's position in classified defense networks. The latter would likely force a reckoning inside the company about whether that outcome was worth the public fight.

The Senate voted 47-53 to reject a war powers resolution sponsored by Sen. Tim Kaine (D-Va.) that would have required the Trump administration to halt military operations in Iran absent congressional authorization. Pennsylvania Democrat John Fetterman was the sole Democrat to vote against the measure; Kentucky Republican Rand Paul was the sole Republican to support it. The resolution represented Congress's first formal attempt to weigh in on Operation Epic Fury, now in its sixth day. Former Senate Majority Leader Mitch McConnell, now chair of the Senate Appropriations defense subcommittee, acknowledged the President has legal authority to act but pressed the administration to articulate a clearer strategic framework, noting that "connecting ends and means is the essence of strategy." A House companion resolution sponsored by Reps. Ro Khanna and Thomas Massie, which calls for an immediate cessation of hostilities, is scheduled for a vote Thursday and is expected to fail. A separate Democratic resolution gives the president a 30-day wind-down window.

Kaine has signaled he will introduce follow-on resolutions as the conflict develops, effectively using the War Powers Act (a 1973 law limiting the executive's ability to engage in armed conflict without congressional approval) as a sustained political lever rather than a one-shot vote. Watch the House vote Thursday closely; the margin and cross-party composition will reveal how durable congressional support for the operation actually is, especially as munitions questions and casualty reports come in over the coming days.

CENTCOM confirmed via social media that the Precision Strike Missile, or PrSM, manufactured by Lockheed Martin, was used in combat for the first time during Operation Epic Fury. The missile was filmed launching from an M142 HIMARS (High Mobility Artillery Rocket System) launcher operating in theater. PrSM Increment 1, the only fielded variant, was designed to replace the aging MGM-140 ATACMS and can strike targets more than 500 kilometers away. CENTCOM did not specify target sets, but known targets throughout the operation have included Iranian command and control centers, ballistic missile and drone sites, and naval assets. The combat debut validates years of Army investment in the program and closes what had been a meaningful gap in long-range ground-fired precision strike following the ATACMS inventory drawdown. CENTCOM also confirmed use of B-2 bombers, F-35s, and the new LUCAS one-way attack drone in the operation.

The Increment 1 combat debut opens the procurement conversation for the follow-on variants that matter most to investors: PrSM Increment 4, for which Lockheed and a Raytheon-Northrop Grumman team are competing on a design capable of flying more than 1,000 kilometers, and a potential Increment 5 configured for autonomous launcher employment at similar range. Combat validation, even from a single engagement, typically compresses acquisition timelines. Watch for post-operation after-action reporting that either confirms or complicates the missile's field reliability.

French President Emmanuel Macron and German Chancellor Friedrich Merz this week formalized a high-level nuclear steering group under the 2019 Treaty of Aachen, establishing a bilateral framework for doctrinal consultation and strategic cooperation. The agreement includes, for the first time, German conventional participation in French nuclear exercises and joint visits to strategic sites. Macron simultaneously delivered a major address announcing an expansion of France's nuclear stockpile and explicitly named Germany, Poland, the Netherlands, Belgium, Greece, Sweden, Denmark, and the UK as potential beneficiaries of an extended French deterrent umbrella. The joint declaration was careful to frame the arrangement as additive to NATO, not a substitute. Significantly, Elbridge Colby, the Under Secretary of Defense for Policy, publicly endorsed "a greater European complexion to NATO nuclear deterrence" at a Council on Foreign Relations event this week, providing rare explicit U.S. backing for the initiative.

For defense industry, the implications extend well below the nuclear threshold. The steering group explicitly encompasses early warning, air defense, and deep precision strike, which are domains where European procurement gaps are well-documented and where the current conflict in Iran is highlighting the cost of chronic underinvestment. Companies competing in European SHORAD (Short-Range Air Defense) and long-range strike will find that a Franco-German strategic architecture is a more reliable procurement driver than NATO aspirations alone. Watch whether Poland, which was named in Macron's speech, formalizes any connection to the framework, given Warsaw's ongoing high-volume defense modernization.

✈️ Air

🛡️ Land

🌊 Sea

🧠 C2 & Cyber

🏭 Industry & Policy

That's the brief. If this landed in the right inbox, forward it on.