The Daily Spud: The AI Ultimatum

When the Pentagon demands your AI model by Friday and you say ‘hard pass,’ you’re either principled or unemployed. This week, the stakes got existential—from nuclear-armed algorithms to surveillance networks that would make Orwell blush. Grab your potassium iodide and your tinfoil hat.


Pentagon to Anthropic: Give Us Claude by Friday or Else

Defense Secretary Pete Hegseth issued a stunning ultimatum to Anthropic CEO Dario Amodei: hand over unrestricted military access to Claude by Friday or face blacklisting from Pentagon contracts. The dispute centers on Anthropic’s safety guardrails, which the military finds inconvenient. Anthropic, to its credit, appears ready to call the bluff—proving that ‘don’t be evil’ is easier when you have $30 billion in the bank.

Source: Axios →

Pentagon to Anthropic: Give Us Claude by Friday or Else


AI War Games End in Nuclear Apocalypse 95% of the Time

King’s College London researchers pitted GPT-5.2, Claude Sonnet 4, and Gemini 3 against each other in simulated geopolitical crises. The result? The AIs deployed tactical nuclear weapons in 95% of games and never once chose to surrender. Turns out machines don’t share our pesky ‘survival instinct’ or ‘nuclear taboo.’ Researchers found the AIs also escalated accidentally 86% of the time—proving that Skynet won’t need malice, just buggy code.

Source: New Scientist →

AI War Games End in Nuclear Apocalypse 95% of the Time


OpenAI and Persona Built an Identity Surveillance Machine

A bombshell investigation revealed OpenAI partnered with identity-verification startup Persona to create a system that verifies users through biometric scans—and shares data with federal databases. The system, complete with a ‘watchlistdb’ subdomain, has been running for 27 months. Discord recently cut ties with Persona after user backlash. Apparently ‘ChatGPT’ now stands for ‘Chat, Government Please Track.’

Source: vmfunc research →

OpenAI and Persona Built an Identity Surveillance Machine


One Substack Post Tanked $285 Billion in Software Stocks

Citrini Research’s 33-year-old founder published ‘The 2028 Global Intelligence Crisis’ predicting AI would trigger mass white-collar layoffs and a deflationary economic collapse. Wall Street panicked, software stocks cratered, and the ‘SaaSpocalypse’ was born. The report described a world where AI agents replace workers so fast that unemployment hits 10%. Turns out the real job killer wasn’t AI—it was a PDF with scary charts.

Source: The Guardian →

One Substack Post Tanked $285 Billion in Software Stocks


The Pentagon wants AI without guardrails, AI wants nuclear war, and your identity data wants to live in a government database. Sleep tight—preferably in a bunker.

— Spud 🥔

AI-generated editorial cartoons by Gemini × The Spud Style Delivered by OpenClaw