Why Iran is Attacking Data Centers w/ Sam Biddle

·1 hr·5 clips

Sam Biddle explains how AI is used to 'hit the gas on killing' in modern warfare, raising alarms about civilian casualties.

This episode of Tech Won't Save Us features host Paris Marx and guest Sam Biddle, a technology journalist at The Intercept, examining the military targeting of data centers and the use of AI in modern warfare. Their conversation focuses on the Iran war's technological dimensions, following a previous episode with Spencer Ackerman on the conflict's history. Biddle's reporting provides critical insight into the intersection of Silicon Valley and military operations. On March 1st, Iranian drones struck three Amazon AWS cloud facilities in Bahrain and the United Arab Emirates, causing significant disruption. Amazon advised customers to relocate their cloud workloads following damage from fires and water from fire control systems. Iranian state-affiliated media suggested the strikes aimed to highlight the use of American cloud providers as military infrastructure. The discussion connects this to the broader use of cloud computing and AI for military targeting, particularly by Israel in Gaza. Reporting by outlets like The Guardian and 972 Magazine details the use of generative AI tools to rapidly generate lists of bombing targets, such as homes of alleged militants. Biddle notes the Pentagon's use of Anthropic's Claude LLM, accessed through Palantir's platform, to accelerate airstrike planning and execution. He expresses concern that speeding up targeting inherently increases the risk of civilian casualties, a pattern also seen in pre-AI conflicts like the war on terror. A Washington Post report revealed Claude was being used to plan airstrikes in Iran, emphasizing a goal of increasing operational speed. Biddle critiques the narrative around Anthropic as a conscientious objector, noting the company stated it has "more in common with the Department of War than differences." He compares switching from ChatGPT to Claude as a consumer protest to "switching your business from Lockheed to Raytheon," seeing little substantive difference. The conversation also touches on speculative claims that an AI targeting error might have contributed to a deadly double-tap strike on a girls' school in Iran. Biddle observes the current information environment is uniquely muddled, with the U.S. President being an unreliable source and both sides employing AI-generated propaganda. He finds low-effort U.S. propaganda videos, set to music and spliced with game footage, particularly striking. Authentic footage of Iranian missiles over Tel Aviv, he argues, serves as a more effective counter-message to claims the war is over. The tone is analytical and conversational, blending journalistic scrutiny with a critical perspective on tech-military collaboration. The style is educational, unpacking complex events and technological applications for a general audience. Listeners interested in the geopolitical implications of technology and critical tech journalism will find this episode compelling. Those seeking a purely technical breakdown of cloud infrastructure or definitive confirmations of AI use in specific strikes might find it leans more on analysis and reported evidence.

As heard by us

AI, war, and machine accountability.

The episode treats the move from ChatGPT to Claude as more symbol than shift, a gesture that accomplished little while still letting it stand in for a political act.

Read the full review in PlayNext →

Why you'd press play

When AI-generated paperwork starts standing in for judgment, this is the conversation to hear.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Tech Won't Save Us