Anthropic vs. the Pentagon: An Imperial Marketing Stunt

In the latest issue of Le Monde diplomatique, I revisit the conflict that pitted Anthropic against the Pentagon, in an attempt to analyze this episode and go beyond the dominant media narrative.

The dispute has all the hallmarks of a hit series: a CEO – Dario Amodei – a freedom-loving idealist standing up to the tyrants of the Trump administration. Except that Anthropic’s “ethical” posture appears totally unwarranted, particularly given Claude’s deployments within Palantir’s systems to accelerate the American kill-chain. Anthropic’s focus on the edge case of autonomous weapons in its negotiations with the US military looks more like a distraction from the enormous stakes raised by the integration of LLMs into military operations.

As for the other “safeguard” put forward by Amodei – the ban on using Claude to mass-analyze data on US residents purchased from data brokers – it left aside a great many other illegal uses of US intelligence surveillance systems.

In the end, the smokescreen appears to have worked, given the commercial and reputational success Anthropic has enjoyed in the wake of this episode. Which in turn also serves the geopolitical interests of the United States. As I explain in the article’s conclusion, at the end of the day,

it is still American models that capture the lion’s share of global markets – vast data sinks all hosted on Amazon, Google, or Microsoft infrastructure, underwriting Uncle Sam’s hegemony.

By playing the “ethical” card, Anthropic seems to be learning from the precedent of the “Snowden controversies” I analyzed in previous research. Caught between their obligations to cooperate with the security state and the reputational cost of public association with its abuses, companies like Google, Apple, Microsoft and their peers had learned not only to collaborate quietly with the government, but above all to stage their resistance in order to reassure their users and shield themselves from the commercial fallout of scandals. Through these hybrid and shifting “security assemblages” – held together by cross-socialization and converging interests, simultaneously transformed and reinforced by the occasional disputes between the actors involved – these companies participated in co-producing the state’s surveillance infrastructures while maintaining a semblance of independence from the government.

The same logic applies today: the “ethics” card allows Anthropic to build a commercial reputation while distinguishing itself from the rest of the tech oligopoly, which – beyond the OpenAI – now eagerly falls in line with the Pentagon – and this despite the timid internal resistance of tech workers (the most notable being that of UK Google DeepMind employees).

Also in this month’s issue of the “Diplo”: an article by Thomas Jusquiame on the astounding profits anticipated by European finance and high-tech companies in light of military “rearmament” and the public spending it entails, and another by Frédéric Lordon on the coming financial crisis – which also underscores the central role of the AI ecosystem in this global house of cards on the verge of collapse.