Pentagon: Grok Gov used in Epic Fury strikes
Confirmed
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 15 November 2026
Your AI and this story
- GPT-6.2 Astra15 days after its cutoff
- Claude Opus 5.5In its training data
- Gemini 4 Hydrogen76 days after its cutoff
- Grok 4.715 days after its cutoff
It is after the cutoff of GPT-6.2 Astra, Gemini 4 Hydrogen and Grok 4.7, and can be in the training data of Claude Opus 5.5.
Key facts
- Sworn statement (quoted by Anadolu, June 17): ‘MSS (Maven Smart Systems) frontier workflows enabled US forces to deploy over 2,000 munitions to 2,000 distinct targets within 96 hours during Operation Epic Fury’; he credited the ‘greatly increased operational efficiency’ of the Grok Gov model; Grok was described as one of four models ‘currently capable of supporting national security applications’ (IBTimes UK)
- Data centres as munitions: ‘The critical ability to field its data facilities at massive scale is as foundational to our modern defense posture as traditional munitions production’; Pentagon network users process about 1.5 billion words a day for logistics, predictive analytics and planning (Anadolu)
- Purpose: the government argued that shutting the turbines that power Colossus 2 would harm national, economic and energy security
- Not stated: what Grok did (target lists, ranking options for analysts, intelligence processing), or whether it selected targets or authorised strikes; the Pentagon has not said which models were used in which strikes
- Context: Maven had relied on Anthropic’s Claude before the Pentagon moved to other vendors (see Pentagon tells BBC it has stopped using Anthropic’s Claude); the Minab school strike in the same campaign is under review (Pentagon review: overreliance on Palantir’s Maven AI contributed to the US strike on a school in Minab, Iran)
- Later: the court denied xAI’s motion to dismiss on Oct 30 (Federal court denies xAI’s motion to dismiss the NAACP’s Clean Air Act suit over gas turbines powering Colossus 2); the claim recirculated in press on Nov 14 (Briefs Finance)
What happened
The declaration became public through a court filing, not an announcement. We have not read the declaration itself; the wording here is from Anadolu’s report, which quotes it, and other outlets agree on the 2,000-targets-in-96-hours figure. What Grok’s part was remains unstated. We found it on Nov 15, 2026 when a Nov 14 article recirculated it; it had not been in the dataset.
Why it matters
It is a sworn statement by a senior US defence official that a commercial chatbot-family model was part of the workflow behind a large strike campaign, the most specific such claim we have found. It also shows the government using that military role in a civil environmental case.
Sources
4 sources from 4 sites. Numbers match the chips in the text.
4 sources: 4 press
Press
- The Hill: Pentagon AI chief: Musk’s Grok chatbot used to launch thousands of missiles at Iran (June 2026)thehill.com, press
- Anadolu: Pentagon reveals Musk’s Grok AI used in Iran waraa.com.tr, press
- IBTimes UK: US reportedly dropped ‘2,000 munitions at 2,000 distinct targets’ in Iran using Grok (Aug 5)ibtimes.co.uk, press
- Task & Purpose: Pentagon AI and Groktaskandpurpose.com, press
Changes
- Filed. Late (news agent)