2026-09-30 — Pentagon announces an Autonomous Warfare Command

• Defense Secretary Pete Hegseth announced a new four-star Autonomous Warfare Command to build and field drones, robots and other autonomous systems across the US military, the first new US command since Space Command in 2019 He said it will have powers similar to the services, be set up over several months and launch within a year, with an interim effort called Project Agincourt and new job specialties for operators of autonomous systems; he spoke at Quantico in his State of the Force speech. (The Straits Times, The Next Web)
• Hegseth tapped Elon Musk, Anduril co-founder Palmer Luckey and former House Speaker Newt Gingrich to lead “Project Meridian,” a 120-day review of the weapons, technologies and capabilities the US military will need in future wars He said the effort “is not about developing new strategies or new policies” but about identifying “the domains that we must conquer, and capabilities we must master”; CNBC notes that Musk’s SpaceX and Luckey’s Anduril both hold large Pentagon contracts. (CNBC, The Hill via Yahoo)

2026-09-25 — A divided D.C. Circuit panel lets the Pentagon keep Anthropic designated a supply-chain risk

• A divided D.C. Circuit panel ruled 2-1 on Friday that the Pentagon may keep its second supply-chain-risk designation of Anthropic in place, so the block on Claude in defence systems can continue The majority wrote that the department “had ample support for its conclusion” that integrating Claude “presented a statutorily covered national-security risk,” and rejected Anthropic’s due-process and free-speech claims. Judge Rita Lin’s August ruling against the other designation stands. Anthropic said it is “considering all options,” which could include asking the full court or the Supreme Court to review it. (Wired)

2026-09-18 — Pentagon investigators cite overreliance on Palantir’s Maven in the Minab school strike

• Pentagon investigators found that flawed intelligence, outdated imagery, a compressed targeting timetable, cuts to civilian-protection staff and overreliance on the AI in Palantir’s Maven Smart System contributed to the February 28 missile strike that killed at least 123 children at an elementary school in Minab, Iran, Bloomberg reported — officials said some Central Command personnel expected Maven to flag stale information in the underlying intelligence; Palantir said there is no evidence its software was at fault and that it is not responsible for the underlying data. The full investigation report has not been released. (Bloomberg)

2026-09-11 — The blacklist is now a fact on the ground, not just a statement: DOD says 90% of classified AI workloads are already off Anthropic, rest done by October

• Under Secretary of Defense for Research and Engineering Emil Michael told reporters at the National Defense Industrial Association’s Emerging Technologies for Defense Conference on 2026-09-11 that the Pentagon is close to completely migrating all classified military AI systems away from Anthropic models: “I’d say about 90% has transitioned. All of the Maven Smart Systems or Palantir work has been transitioned months ago, and we’re on track to get it done by the end of the month.” This is a concrete operational follow-through on the 09-03 entry below (Michael’s public statement that Anthropic “is still a designated Supply Chain Risk”), naming the replacement vendor set for the first time: SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, AWS and Oracle. Michael framed the multi-vendor shift as a lesson independent of the Anthropic dispute itself — “one of the mistakes… is the fact that the Pentagon was solely reliant on one model” — while confirming the two sides are still litigating the underlying designation in court. (DefenseScoop)

2026-09-03 — Pentagon says Anthropic blacklist stays in effect, defying the court and Lutnick’s truce signal

• Pentagon Under Secretary of War Emil Michael said Anthropic “is still a designated Supply Chain Risk” and remains one “for the Defense Industrial Base,” a public statement that directly contradicts both Judge Rita Lin’s August 27 permanent injunction voiding the designation and Commerce Secretary Howard Lutnick’s own 09-02 remarks that Anthropic is “back on the right side” and the administration “trust[s] Anthropic.” The statement lands one day after this map recorded Lutnick’s reconciliation framing, and shows the administration is not unified on it — Defense is maintaining the blacklist Commerce says is resolved. No notice of appeal of the August 27 ruling has been filed as of today, per Washington Examiner’s explicit check of the record; separately, the same ruling’s 7-day administrative-stay question may itself need correcting (see Ledger checks above) — either way, the Pentagon’s practical posture as of today is continued non-enforcement of the court’s order rather than compliance. (Axios, Washington Examiner, Bloomberg)

2026-08-31 — Pentagon adds ChatGPT and Grok to its military AI portal; Claude stays excluded

• The Pentagon added OpenAI’s ChatGPT Mil and xAI/Starshield’s Grok for Government to GenAI.mil, its secure internal AI portal, alongside the Google Gemini access it already offered, effective Monday 2026-08-31 — both new models are accredited for Controlled Unclassified Information at Impact Level 5, and DoD said the point is giving warfighters “distinct technical strengths of each top-tier large language model.” The portal had already onboarded 1.7 million of the Defense Department’s roughly 3 million military, civilian and contractor personnel as of this expansion. (DefenseScoop, TechCrunch, Military Times, DoD primary)
• Claude is still absent from GenAI.mil, six days after Judge Rita Lin ruled the Pentagon’s “supply chain risk” designation on Anthropic unlawful (the ruling Frontier Gatekeeping and this thread both already carry). DefenseScoop’s account restates why: Anthropic held a contractual line against Claude being used for mass domestic surveillance or fully autonomous lethal weapons, the Pentagon “demanded unrestricted access” instead, and the designation followed — a court win for Anthropic on the legal question has not (yet) translated into procurement access. ⚠️ A single outlet (Tech Times) additionally reported that xAI’s own engineers found no reliable fix for Grok’s CSAM-generation problem before this deployment, sourced to an unnamed “contemporaneous internal investigation” — I could not independently verify this (the article 403’d on direct fetch and no second outlet was found corroborating the internal-document claim specifically), so it is noted here only as an unconfirmed lead, not a finding. (DefenseScoop)

2026-08-27 — A federal judge ruled the Pentagon’s “supply chain risk” label on Anthropic unlawful; no appeal has been filed (heading corrected 2026-09-02)

• U.S. District Judge Rita Lin ruled Thursday evening, 2026-08-27, that the Pentagon’s designation of Anthropic as a “supply chain risk” was unlawful on three separate grounds: First Amendment retaliation against the company for its public safety positions, a violation of the Administrative Procedure Act (“arbitrary and capricious”), and denial of the pre-deprivation process required under the Fifth Amendment. Lin specifically noted the government’s own position was self-contradicting — labeling Anthropic a security threat while simultaneously proposing to invoke the Defense Production Act (which designates firms as essential to national security) and continuing to pursue contracts with it. This resolves, in Anthropic’s favor, the skepticism this thread logged at the 08-23 hearing. (CNN, TechCrunch, NPR)
• The ruling took effect on 08-27 with no stay of any kind, and no notice of appeal had been filed as of 09-03. ✏️ Corrected 2026-09-03: this line previously read “The ruling carries a 7-day stay, expiring 2026-09-03.” The Order of Final Relief (Dkt. 251, read from the RECAP PDF) contains no stay language — the words “stay,” “seven” and “appeal” do not appear in it — and the docket has no separate stay order; Decrypt (08-28) and the Washington Examiner (09-03) report Judge Lin denied the government’s request for a seven-day administrative stay. The docket still ends at Dkt. 252 Judgment and “Civil Case Terminated” (08-27). This is the second correction to this entry in two days; the dated expectation built on the stay (anthropic-dow-stay-expiry) was withdrawn the same day. (Order of Final Relief, Dkt. 251, CourtListener docket, Washington Examiner) ⚠️ Corrected 2026-09-02: this entry previously read “and the Defense Department has already filed a notice of appeal to the Ninth Circuit.” No such filing exists. A live check of the N.D. Cal. docket (3:26-cv-01996-RFL) shows the last entries are the August 27 order, order of final relief, judgment, and “Civil Case Terminated” — no notice of appeal after judgment, and no new Ninth Circuit case tied to this ruling. The only Ninth Circuit docket in this dispute (26-2011) is the earlier interlocutory appeal of the March preliminary injunction, filed 2026-04-02 and stayed 2026-04-27 pending the separate D.C. Circuit FASCSA case, still dormant. Re-checking the three sources this entry cited: NPR (08-28) says only that a source “thinks they might appeal”; TechCrunch (08-28) does not mention an appeal at all. The original entry appears to have conflated the March/April stay-and-appeal cycle — which did produce a real notice of appeal and a quoted Emil Michael statement — with the August 27 ruling, which has not. Read as still open until the stay lapses 09-03, not as an appeal already filed. This also resolves an internal contradiction: Frontier Gatekeeping has recorded since 08-29 that “no notice of appeal has actually been filed.” Background: Hegseth designated Anthropic a supply chain risk in February 2026 after the company refused to remove guardrails preventing military use of Claude for autonomous weapons/mass surveillance; Anthropic sued in March; this is the first of two suits (California, decided; a separate DC suit remains open). No Ninth Circuit ruling has landed as of 09-01 — still open, not resolved. Anthropic’s stated response: it “welcomes the court’s ruling” and remains “focused on working productively with the government.”

2026-08-23 — A federal judge signals skepticism at the Pentagon’s “supply chain risk” label on Anthropic

• A federal judge signalled skepticism at the Pentagon’s “supply chain risk” designation on Anthropic, at a hearing following Saturday’s temporary block of the label — which had been restricting federal contractors from using Claude. The judge reportedly pushed back hard on the government’s arguments in terms suggesting the designation is unlikely to survive. ⚠️ Reduced confidence, stated on its face: the article page 404’d on fetch, so only the RSS headline and pubDate are confirmed and the body of this account is unverified. Kept rather than dropped because the hearing is corroborated by the prior day’s order from the same outlet (Saturday’s temporary block). Re-check the full article before anything is built on this entry. (AP via ABC News)

2026-08-15 — The Pentagon’s data-platform rename draws scrutiny over whether it’s real consolidation

• DefenseScoop reported doubts that the Pentagon’s “War Data Platform” — the rebrand of the former Advana system, the Chief Digital and AI Office’s common data foundation meant to feed AI tools across the military — actually fixes what plagued the original program, after Accenture Federal Services was awarded an up-to-$821M task order (via the CDAO with GSA as acquisition partner) to lead the integration effort. A former senior defense official quoted in the piece: “You don’t fix a broken program by changing the name on the cover sheet and then hiring the same type of firm, and many of the same characters, that produced the original problems” — a direct hit on this thread’s core question of whether “consolidation” is structural or cosmetic. (DefenseScoop)