Palantir CEO Alex Karp took the stage at Andreessen Horowitz's American Dynamism Summit on Friday to deliver a familiar but sharpening argument: the United States must be willing to build and deploy AI weapons, or cede that capability to adversaries who will. He named Iran specifically as the scenario that makes the stakes concrete.
Karp has made versions of this argument for years, but the a16z forum gave it fresh amplification. On Iran, he argued that U.S. military and intelligence agencies need AI-enabled targeting and analysis tools capable of operating faster than any human decision loop — and that Palantir's platforms are among the few commercially built systems already embedded in those workflows at scale. He described reluctance among tech companies to do defense work as a form of strategic self-harm dressed up as ethics.
The two platforms doing that work are Gotham, which handles intelligence analysis for U.S. defense and intelligence agencies, and AIP — <a href="/news/2026-03-14-palantir-demos-show-how-the-military-could-use-ai-chatbots-to-generate-war-plans">Palantir's AI Platform</a> — which deploys large language models in operational defense contexts. AIP is the closer fit for what Karp was describing: a system that ingests intelligence, surfaces recommendations, and can feed directly into command decisions. That's not a chatbot. It's an agent stack running in environments where the outputs have lethal consequences.
That's the detail worth sitting with for anyone covering the AI agent space. Palantir is not building toward agentic AI — it's already operating it, at classified scale, in active military and intelligence use cases. Most of the agentic AI discussion in the industry centers on enterprise productivity and software development. Karp's summit appearance was a reminder that the highest-stakes deployment of autonomous AI systems is happening in a domain where most of the industry has <a href="/news/2026-03-14-pentagon-anthropic-claude-military-red-lines">chosen not to compete</a>.
Andreessen Horowitz has been deliberate about building that competitive pressure. The American Dynamism thesis — the firm's framework for investing in defense, manufacturing, and national infrastructure — has drawn a cohort of founders and executives who share Karp's view that Silicon Valley's reflexive avoidance of government work was a mistake. Palantir, which went public in 2020 and has posted consecutive profitable quarters since 2023, is the clearest proof-of-concept for that bet.
The a16z summit didn't produce a policy announcement or a contract disclosure. What it produced was Karp, on a prominent stage, arguing that the ethics of AI weapons are downstream of whether the right side wins. His next test of that argument will likely come in front of a different audience: Palantir is currently competing for several major DoD AI contracts expected to be awarded in the second half of 2026.