Anthropic vs The Department of War: Opportunities for the EU? (Copy)

By Laurent Ledoux

Can Europe seize the moment and deepen its partnership with Anthropic?  What does the fight between Anthropic and the Pentagon tells us about Conscious leadership?

In this article I will investigate two related questions: Could the EU and/or big European (Tech) companies strike a mutually beneficial deal with Anthropic, following Trump and Hegseth's recent decisions regarding Anthropic? And what does all this teach us about conscious leadership? While so many self-titled world leaders in politics or business are grotesque clowns with no moral compass, can the founder and CEO of Anthropic, Dario Amodei, be seen as a conscious leader?

What just happened?

Trump ordered all federal agencies to immediately stop using Anthropic's technology, and Hegseth designated Anthropic a "supply chain risk to national security" — a category normally reserved for companies from adversarial nations like China. The core dispute: Anthropic refused to drop two red lines: (1) no use of its AI for fully autonomous weapons and (2) no mass domestic surveillance of American citizens.

The kicker? OpenAI swooped in hours later with its own Pentagon deal, even though Sam Altman said his company has the same "red lines" (the fundamental difference seems to be whether these lines are included in the model—as requested by Anthropic—or in the contract—as requested by OpenAI: this obviously does not provide the same guarantees). Anthropic says for its part it will challenge the supply chain risk designation in court.

What could this mean for Europe?

Anthropic was already deepening its European footprint before this blew up. EMEA is Anthropic's fastest-growing region, with revenue growing over 9x in the past year, and the company recently opened offices in Paris and Munich alongside existing ones in London, Dublin, and Zurich. They struck an enterprise deal with German insurance giant Allianz in January, and their European client list includes BMW, SAP, L'Oréal, and Sanofi. Anthropic also signed the EU's General-Purpose AI Code of Practice, praising its transparency and safety principles.

Meanwhile, there's a broader EU push toward what's being called "EuroStack" — an initiative to replace dependence on American tech with European alternatives for AI, cloud computing, and hardware, driven partly by fears that Europe's tech dependency could be weaponized by the Trump administration.

Has anyone specifically proposed an EU-Anthropic deal post-blacklist?

Not yet — at least not publicly. This only happened hours ago, and EU officials have so far only "expressed concern about the potential fragmentation of global AI ethics standards" rather than making concrete moves. European political cycles being what they are, it'll take days to weeks before you'd see formal proposals.

But could it happen? Here's my take.

The alignment is almost uncanny. Anthropic's two red lines — no autonomous weapons, no mass surveillance — are basically the EU's values written in AI company letterhead. The EU AI Act already restricts many of the exact things Anthropic is fighting about. So on a values level, this is a natural partnership.

But there are real complications:

1. The "supply chain risk" threat. If Hegseth's designation sticks (Anthropic disputes its legal validity), it would bar any company that does business with the U.S. military from conducting commercial activity with Anthropic. For European defense contractors who also work with the Pentagon — think Airbus Defence, BAE Systems, Thales — that's a serious chilling effect. European companies would face a choice between Anthropic and their U.S. military business.

2. Anthropic's corporate structure. Amazon holds a massive minority stake, and Nvidia and Microsoft committed up to $15 billion in investment last November. Anthropic is deeply embedded in the American tech ecosystem. The EU couldn't just "adopt" it the way you might bring in a company that's looking for a new home base.

What could work. The more realistic play isn't some grand sovereign partnership — it's a deepening of what's already happening. European governments could accelerate adoption of Claude in their own public services (where the U.S. supply chain designation likely doesn't apply). Major European enterprises could signal support by expanding their contracts. And the EU could position itself explicitly as a regulatory environment where companies that take safety seriously are rewarded rather than punished.

However, let us be clear about what this means — and does not mean — in terms of sovereignty. Anthropic's values may resemble the European AI Regulation, but Anthropic remains a for-profit company valued at £380 billion, expanding across Europe on American cloud infrastructure. Replacing dependence on one American company with dependence on a ‘slightly more virtuous’ American company is not sovereignty. It's the same game with a different flag. Principles enshrined in a corporate mission statement are only sustainable until the next round of funding or the next change in administration. This is not bad faith towards Anthropic — it is a matter of structural reality. Therefore, beyond possibly deepening ties with Anthropic, the EU should above all use this moment to accelerate what Dr Cecilia Rikap's Coalition for Democratic and Ecological Digital Sovereignty is proposing: democratically governed AI commons, built with public funds, that no political reversal in Washington could revoke. Willy Debacker explored this issue in detail in his article: Digital Sovereignty: Taking Back Control. That is the real challenge: not simply choosing a better American supplier, but building an alternative that belongs to European citizens.

Meanwhile, the Accenture report on sovereign AI found that 73% of European organizations want governments and the EU to play a key role in enhancing digital sovereignty through regulations, subsidies, or public investments. A McKinsey analysis estimated sovereign AI could unlock up to €480 billion in annual value for Europe by 2030. The political appetite and economic logic are there.

In conclusion, the conditions for an EU-Anthropic deal are remarkably ripe. Anthropic's values, its existing European expansion, and the EU's growing urgency around digital sovereignty all point in the same direction. The real question is whether European leaders will have the courage to go beyond the partnership with Anthropic and invest heavily in public and democratic AI commons, and whether they will act quickly enough (which, historically, has not been their strong point) before the narrative becomes set in stone.

This is one of those moments where the geopolitics of AI just got very, very real — and Brussels has about a two-week window to say something meaningful before the narrative hardens.

Now, since this newsletter is called “The Conscious Leader”, I do not want to miss this opportunity to address what conscious leadership is about and share with you the reflections that were triggered by another question.

Can the current actions of Dario Amodei be considered those of a conscious leader?

What just happened is basically a live stress test of whether someone's leadership philosophy holds up when the cost gets real. "Conscious leadership" gets thrown around a lot in business circles, often as vague branding. But when the concept has teeth, it usually means something specific:

A conscious leader, is a leader who makes decisions with explicit awareness of their broader impact (on stakeholders beyond shareholders, on society, on long-term consequences) and who's willing to absorb personal and organizational cost to stay aligned with those principles.

By that definition, what Amodei just did is a pretty remarkable case study.

1. He didn't just talk about values when it was cheap. Anthropic has been saying since its founding that AI safety isn't a marketing slogan: it's the reason the company exists. The Amodei siblings left OpenAI over exactly these kinds of directional disagreements. But lots of founders say principled things during fundraising rounds. The question is always: what happens when the principle costs you something? And here, the cost is enormous. We're talking about losing all U.S. federal contracts; being designated a supply chain risk, a category normally reserved for Huawei-type adversaries; having the President of the United States publicly call you a "leftwing nut job" right as you're preparing an IPO; Anthropic's valuation is $380 billion. The personal financial exposure for Amodei is staggering.

2. He was specific about his red lines, not vague. This matters. He didn't say "we care about ethics" in some hand-wavy way. He drew two precise boundaries: no mass domestic surveillance of Americans, and no fully autonomous weapons without human oversight. And he explained why: not on ideological grounds, but on practical ones. Current AI models aren't reliable enough for autonomous lethal decisions. That's an engineering argument wrapped in an ethical stance. It's harder to dismiss than pure moralizing. [Side note: Some commentators argue that these safeguards already exist. They cite, in particular, Department of Defence Directive 3000.09, which requires ‘appropriate levels of human judgement’ for the use of force and a thorough review before the development of autonomous weapons. The argument is worth hearing, and the question of whether a private contractor should be able to dictate the conditions of use to the Pentagon is a legitimate one. But upon examination, Directive 3000.09 is a governance framework, not a prohibition. It authorises autonomous weapons subject to an approval process — which the Deputy Secretary of Defence can even bypass for ‘urgent military need’ (Section 4.2). Several categories of autonomous weapons do not even require senior review: facility defence, defence of remotely piloted vehicles, non-lethal force against material targets. It is a directive, not a law — issued under Kathleen Hicks in January 2023, and subject to change by any new administration. Calling it a ‘red line against fully autonomous weapons’ is misleading. Most importantly, the Directive says absolutely nothing about mass surveillance. It only addresses weapons systems. Anthropic's argument is precisely that AI enables forms of surveillance that are technically legal but fundamentally incompatible with fundamental rights: large-scale automated analysis of commercial data, entity resolution between databases, predictive risk scoring, continuous behavioural analysis. The government could purchase commercial data sets and analyse them with AI without technically falling under the scope of illegal surveillance. This is a legal blind spot that Directive 3000.09 does not cover — and which Amodei explained in detail in his interview with CBS News.]

3. He left the door open. His public statements made clear Anthropic wants to work with the military on everything else. He offered to help with the transition to other providers if the Pentagon chose to walk away. That's not the behavior of someone grandstanding; it's someone trying to hold a line while minimizing unnecessary conflict. Whether you think he misjudged the politics is a separate question.

4. The workforce followed. Within hours, hundreds of employees at Google, OpenAI, Microsoft, and Amazon were signing petitions asking their own companies to adopt the same red lines. OpenAI's Altman publicly said he shares Anthropic's position. That kind of ripple effect doesn't happen unless people perceive the stand as genuine rather than performative.

Now... here's where there could be a push back on an uncritical "yes, he's a conscious leader" answer.

1. There's a tension in Anthropic's position that's worth naming. The company built Claude for classified military operations. It was used in the Venezuela raid that resulted in 83 deaths. Anthropic chose to be deeply embedded in national security work. So the "conscious" framing gets complicated — this isn't a pacifist company that stumbled into a Pentagon contract. It's a company that said "we'll do the hard, morally complex work of military AI, but with these two guardrails." Some critics would say that's not consciousness, it's just drawing an arbitrary line in a much larger ethical gray zone.

2. And there's also a pragmatic reading: Anthropic may have calculated that its principled stand would generate enormous goodwill (which it has), boost its brand for enterprise customers who care about trust (particularly in Europe), and ultimately be worth more than the federal contracts it's losing. Conscious leadership and strategic brilliance aren't mutually exclusive... but they're also not the same thing.

In conclusion, what is my assessment? I think that, in this moment, Amodei is closer to the real thing than most tech leaders who claim the label. The founding story checks out — he left a powerful position at OpenAI over genuine disagreements. The company's structure (the Long-Term Benefit Trust, the PBC status) suggests institutional commitment, not just personal charisma. And the willingness to absorb this level of punishment rather than quietly capitulate... that's rare.

But "conscious leader" is a description and a work in progress, not a verdict. Let’s follow this closely and consider it again in six months — after we see how the court challenge plays out, what happens with the IPO, and whether the red lines hold if the pressure gets even worse.

In the meantime, hats off for Amodei. In such dark and troubled times, he gives us all an example of what it means to stand up against autocratic bullies.