Science Weekly Does Trump want to wage an AI-powered war?
17 snips
Mar 3, 2026 Chris Stokel-Walker, tech journalist and author on AI and policy, discusses the militarisation of artificial intelligence. He outlines Anthropic’s safeguards and the Pentagon split, the swift pivot to OpenAI, and how AI is reportedly shaping recent strikes. The conversation explores why this may mark a dangerous turning point for future conflicts.
AI Snips
Chapters
Transcript
Episode notes
Anthropic's Contract Red Lines
- Anthropic imposed two red lines when contracting with the Pentagon: no autonomous weapons and no mass domestic surveillance.
- Anthropic limited Claude to lawful foreign intelligence and counterintelligence use because it deemed current models unreliable for lethal autonomy.
Trump Ousted Anthropic And Brought In OpenAI
- Donald Trump publicly declared Anthropic persona non grata after they resisted military uses he wanted, ending their government access.
- Within hours OpenAI struck a new deal with the Pentagon to fill the void left by Claude's exclusion.
Contracts May Not Prevent Presidential Override
- OpenAI agreed terms the Pentagon described as similar to Anthropic's, including limits on domestic monitoring and autonomous weapons.
- Practical enforcement may depend on presidential will more than contractual language, Chris warns.