Grok Goes to War: AI Governance Breaks Down in Real Time
How this was made Verified AI
Every Intellegix briefing is generated from that day's broadcast and run through automated checks before it publishes — with a human paged on any flag. Here is the trail for this edition.
The Pentagon's use of Grok — Elon Musk's AI product from xAI — to help plan the Iran strikes, confirmed in a DOJ filing, is a development that deserves treatment as major news rather than a footnote. The implications run in several directions simultaneously. There is a capability question: what role did the AI actually play — generating options, analyzing intelligence, running simulations? There is a conflict-of-interest question: Musk holds significant defense contracts through SpaceX, is a prominent political supporter of the current administration, and his AI product was used for active military strike planning with all the contractual and data access implications that entails. And there is an accountability question that oversight bodies have not yet begun to address publicly.
ChatGPT going live on the Pentagon's GenAI.mil platform in early July — giving three million defense personnel access to the system for sensitive but unclassified work — creates a parallel track that is somewhat less fraught from a conflict-of-interest standpoint but raises its own concerns about data leakage and the breadth of what qualifies as 'sensitive but unclassified.' Trump administration officials simultaneously denied G7 allies access to Anthropic's most advanced AI models, meaning allied military and intelligence establishments cannot legally access the same frontier AI tools that American counterparts are using — a capability gap within the alliance that is counterproductive if the goal is coordinated response to shared threats.
China's Zhipu AI launched GLM-5.2, offering a one-million-token context window under an MIT license, days after the U.S. ordered Anthropic to shut down its Fable 5 model. The timing appears deliberate. An MIT license means any developer globally can use, modify, and deploy the model without restriction — a direct attack on the business model of every commercial AI provider. The U.S. delay in blacklisting DeepSeek and more than one hundred other Chinese firms, intended to preserve diplomatic relationships, extended the window during which those firms continued developing capabilities and international partnerships.
Sixteen thousand SAG-AFTRA members signed a letter pushing Congress to pass the NO FAKES Act ahead of Thursday's Senate Judiciary Committee vote, asking for the first federal law establishing explicit protections against AI-generated likenesses without consent. The entertainment industry's experience — where deepfake technology is already being used commercially in ways actors have not authorized — represents the leading edge of a problem that will eventually affect ordinary people's control over their own identities. Bipartisan support for the bill is substantial, though lobbying pressure from technology companies opposing broad liability provisions is also significant.