Now Playing
3M Military Users, AI Hallucinations in Targeting & VC Bets on Agent Risk
Audio is available on Spreaker — see link below.
Three million Pentagon personnel are using commercial AI models for national security tasks — and a single hallucination in a targeting chain could trigger escalation. Plus, venture capital exits foundation models as domain-specific AI raises signal where defensibility now lives.
Audio is available on Spreaker — see link below.
Three million U.S. military and civilian personnel are now using ChatGPT through a Pentagon platform called GenAI.mil. That's the number.
Here's where it gets specific. An analyst used a chatbot to flag a Chinese ship as a nuclear weapons carrier.
The important distinction here is structural. Pentagon policy already requires that AI systems used in defense contexts be traceable, governable, and understood.
Shift from Washington to the funding landscape, and a different signal emerges. Early-stage capital is moving away from frontier model dependency and toward domain-specific applications.
The Humanos raise is worth noting separately. Autonomous AI agents are gaining real financial and operational permissions.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.