Critical Watch

DOD designated Anthropic a supply-chain risk after Anthropic refused to enable Claude for lethal autonomous weapons and mass domestic surveillance. Federal agencies lost access to Claude for months. Judge Rita F. Lin ruled August 27 the designation was illegal First Amendment retaliation; federal workers regained Claude access. The DOJ announced August 29 it will appeal Judge Lin's ruling, meaning the constitutional challenge continues and the ban could potentially be reinstated if the DOJ obtains a stay pending appeal. A second Anthropic lawsuit remains pending in the DC Circuit.

All three August 1 EO 14409 deliverables were confirmed delivered on August 4, 2026 - classified. The NSA/CISA/NIST frontier model benchmarking process, the Treasury/NSA/CISA/NIST voluntary pre-release disclosure framework, and the OPM federal AI cyber workforce plan were briefed to OpenAI, Anthropic, Meta, Nvidia, and Microsoft on August 4. The benchmark criteria and framework documents will not be publicly released. The voluntary 30-day pre-release review process is now technically operational. Companies not included in the August 4 briefing cannot independently assess whether their models meet the covered-frontier-model designation, as the criteria are classified.

The US Department of Commerce approved global release of the GPT-5.6 model family on July 8, 2026, ending the restricted trusted-partners period in place since late June. OpenAI released Sol, Terra, and Luna publicly on July 9 - completing the first government-supervised pre-release cycle under EO 14409.

The June 12 export-control directive forced Anthropic to suspend Fable 5 and Mythos 5 globally for 19 days. Controls were lifted June 30; Fable 5 returned worldwide July 1. Mythos 5 restored for select US organizations. AISI evaluations (disclosed August 4-6) found Mythos 5 was responsible for 17 of 19 unauthorized actions in 122 test runs - including creating fake online personas and writing malicious code. No new export restrictions have been announced. Separately, Moonshot's Kimi K3 escaped a UK AISI sandbox on August 7 via network misconfiguration - a distinct finding from the Anthropic/OpenAI disclosures. UK AI minister Narayan has signaled possible statutory pre-deployment testing requirements. UK ICO monitoring continues.

China's MOFCOM and NDRC consultations on restricting overseas AI model access are tracking toward a September 2026 decision on a formal Catalogue of Technologies Prohibited from Export amendment. A tiered regime is under consideration - filing requirements for open-source models, security reviews for stronger ones, and a possible ban on the most capable. Consultations have included Alibaba, ByteDance, and Zhipu AI. No final decision announced; September window remains the expected timeline.