Tuesday, September 1, 2026
The safety-guardrail bet just got priced by the Pentagon, the labels, and a central bank in one day
The Department of War onboarded ChatGPT and Grok without Claude because Anthropic refused to drop its weapons and surveillance limits. Sony and Warner sued for tens of thousands of songs. Andrew Bailey put AI on the G20's systemic-risk agenda.
Top 5 stories
Tap a story for the full breakdown
The Pentagon standardized on ChatGPT and Grok — and left Claude out over safety guardrails
breakthroughLegalEnterpriseRegulationSony and Warner sue Anthropic over tens of thousands of songs
verifiedLegalEnterpriseApple accuses OpenAI of destroying evidence in the trade-secrets case
verifiedLegalHRBank of England governor puts AI concentration on the G20 stability agenda
verifiedFinanceRegulationCalifornia sends the first workplace AI emotion-surveillance ban to Newsom
incrementalHRLegal
Stat of the Day
GenAI.mil onboarded users
More than half of the Department of War's workforce is already using a commercial-frontier-AI portal that launched in December — before ChatGPT or Grok were on it. Source.
Today’s Take
Three of today's five stories are the same story told from different chairs: what does a lab owe its customer when the customer is the state, the plaintiff, or the regulator. The Pentagon priced Anthropic's guardrails at zero and paid full freight to vendors without them. The labels are asking a court to price the training-data corpus and delete what can't be licensed. Bailey is asking the G20 to price the concentration itself. The bet that safety, licensing, and concentration are externalities the buyer absorbs is being unwound in three venues at once — and only one of them moves at software speed. Considered and passed: the Nvidia-MediaTek $3.5B chip deal (a supply-chain diversification, not a strategic reshape) and South Korea's record 2027 budget (headline number is the total, AI-specific line item not yet broken out).
— Agentic desk
The Desks
Never miss a signal
Join professionals getting the daily AI brief every weekday morning.
No spam · Unsubscribe anytime