What Next | Daily News and Analysis TBD | If You Give A.I. a Nuke
20 snips
Nov 30, 2025 
Josh Keating, a senior correspondent at Vox and a national security researcher, discusses the integration of AI into nuclear systems. He explores the potential roles AI could play in decision-making chains, from predictive maintenance to automating retaliatory plans. However, Keating highlights military skepticism about ceding control to AI, emphasizing risks like errors and deception. He warns that an AI arms race could escalate geopolitical tensions, reminding us that human judgment remains crucial amidst growing reliance on AI in warfare.
AI Snips
Chapters
Transcript
Episode notes
False Alarm Near-Miss In 1979
- In 1979 Zbigniew Brzezinski almost ordered a nuclear response before discovering a faulty chip caused a false alarm.
- The near-miss shows humans once averted catastrophe by pausing and verifying sensor errors.
Petrov's Decision Prevented Retaliation
- In the 1980s Soviet officer Stanislav Petrov chose to wait after his system falsely flagged a US launch.
- His restraint prevented an automatic retaliatory launch based on a sensor misread of sunlight on clouds.
Low-Tech Systems Meet AI
- Much of the U.S. nuclear command system remains low-tech, with recent modernization long overdue.
- Integrating AI into that ecosystem risks blurring where human judgment ends and machine input begins.