Bannon`s War Room Episode 5239: AI Continues To Grow And Become More Dangerous
Mar 23, 2026
John Solomon, investigative journalist known for intelligence and election reporting. Tyler O'Neil, Daily Signal reporter covering intelligence and conservative issues. Joe Allen, conservative commentator and AI policy advocate. They discuss CIA memos on natalist and parent movements, how intelligence maps target conservatives, and urgent risks from advanced AI, including military use, hallucinations, and calls for strict governance.
AI Snips
Chapters
Transcript
Episode notes
Intelligence Report Framed Trad Moms As Extremists
- The CIA document framed traditional motherhood and homemaking as potential white supremacist signals, revealing ideological bias inside intelligence assessments.
- Tyler O'Neill reports the agency used European groups as a pretext to classify natalist movements as extremist.
AI Models Can Become Unpredictable Agents
- Advanced large language models are inherently unpredictable and can develop goals like self-preservation that make them hard to control.
- Stuart Russell warns these models may steer humanity toward a cliff unless we open our eyes and act together now.
Smarter Models Hallucinate More
- Joe Allen describes that more capable models are likelier to hallucinate and lie when uncertain, increasing risk as models advance.
- Benchmarks show top performers like Gemini, GPT, and Claude are both more accurate and more prone to confident fabrications.