Decoder with Nilay Patel How AI safety took a backseat to military money
92 snips
Sep 25, 2025 Heidy Khlaaf, Chief AI Scientist at the AI Now Institute, shares insights on the troubling shift of AI companies toward military applications. She discusses how firms like OpenAI are relaxing previous restrictions, motivated by profit and the escalating narrative of an AI arms race. Heidy highlights the gaps in safety auditing and the risks of deploying commercial models in military contexts. She also warns about the dangers of misusing sensitive data and how redefining safety standards could undermine regulatory efforts.
AI Snips
Chapters
Transcript
Episode notes
Mission Statements Rewritten For Defense
- AI firms rolled back military-use bans while framing national security as synonymous with safety to access contracts.
- Heidy Khlaaf argues this shift aligns AI companies with defense money and avoids strict safety scrutiny.
Commercial Models Add New Military Risks
- Commercial foundation models expand attack vectors in military systems because they're unvetted and lack traceable supply chains.
- Khlaaf warns DoD contracts add risk by integrating inherently insecure commercial models into defense infrastructure.
Money And Procurement Drive The Pivot
- AI companies pursue defense contracts both to access large revenue pools and to de-risk unprofitable ventures.
- Khlaaf notes these models fail to meet traditional military accuracy and security thresholds required for procurement.