Computer Says Kill: Collapsing the Chain W/ with Matt Mahmoudi

In this opening episode of the "Computer Says Kill" series, host Alex Dunn sits down with Matt Mahmoudi, an assistant professor at Cambridge and Amnesty International advisor, to expose how U.S. defense agencies are already deploying large language models—such as Anthropic's Claude—to compile target lists for operations in Iran, Venezuela, and Gaza. The conversation frames AI‑enabled warfare as a watershed moment where algorithmic decision‑making bypasses traditional human judgment, turning probabilistic outputs into lethal directives. By highlighting concrete examples of AI‑driven kill‑chain integration, the episode underscores the urgency of understanding the technology’s role in modern conflict.

Mahmoudi explains that tech firms are not neutral providers; they monetize the militarization of generative AI, offering contracts that promise efficiency while simultaneously distancing themselves from accountability. The episode details how these models ingest biased, often synthetic data—social‑media posts, iris scans, refugee aid records—producing skewed probability scores that disproportionately target racialized populations. Companies like Anthropic navigate a dual narrative, presenting AI as a defensive tool yet enabling autonomous weapon systems without transparent oversight. This convergence of profit motives, data bias, and normalized rhetoric compresses the traditional kill‑chain, eroding legal safeguards and making it harder to trace responsibility.

The discussion concludes with a call for robust policy interventions. Mahmoudi stresses that human operators must retain meaningful control, and that international humanitarian law requires clear attribution of decisions throughout the AI‑augmented targeting process. Transparency, independent audits, and strict export controls on AI weaponization are presented as essential to prevent a future where algorithmic outputs become the default justification for lethal force. For business leaders and policymakers, the episode offers a stark reminder: without proactive regulation, the integration of large language models into warfare threatens both ethical standards and global security.