The Lyceum: AI Intelligence Brief — May 28, 2026

Two clocks converged in the last 24 hours. The Section 1533 cross-functional team (CFT) mandated by the FY2026 NDAA is T-4 days from its June 1 statutory standup deadline with no public constitution signal; assess with moderate confidence the CFT is either unbuilt or operating without disclosure. Simultaneously, the HASC FY27 NDAA chairman's mark dropped May 26 with full committee markup locked for June 4, meaning Congress is writing the second floor of the DoD AI governance stack before the FY26 foundation is poured. The BIS replacement rule for the rescinded AI Diffusion Framework remains absent at Day 381. The seven-day HASC amendment window is now the only near-term lever to either correct the unexecuted FY26 mandates or codify a congressional export framework before BIS acts.

[INTEGRATION] CDAO Wingman ATO Progressing: CDAO's "Wingman" task-automation platform is moving through its authority-to-operate process via the War Data Platform team, per DefenseScoop. The ATO baseline — not the platform itself — sets the security floor against which PSP-adjacent workflow automation will be measured. PMs evaluating enterprise AI integration should treat this as a live deployment standard, not a pilot.

[INTEGRATION][AUTONOMY] PSP July Demo Clock at T-34: Initial demonstrations for the seven Pace-Setting Projects — including Swarm Forge and Agent Network — are due July 2026, per the DoD AI Strategy memo. No public record confirms DoDD 3000.09 autonomous weapons adjudication has been applied. If demos proceed without it, the Department demonstrates autonomous capability outside its own legal framework.

[ACQUISITION] Xiaomi MiMo Inference Price Cut (up to 99%): (vendor claim — magnitude unverified by independent benchmark) Chinese inference pricing pressure intensifies. Acquisition leads using commercial inference cost models in trade studies should refresh assumptions; the directionality is high-confidence even if the headline number is not.

Section 1533 of the FY2026 NDAA requires SECDEF to stand up a CFT for AI model assessment by June 1, 2026, with the assessment framework due June 1, 2027. At T-4 days, no public standup notice exists. The decision implication: every AI acquisition adjudication in the Department this week operates against no authoritative DoD-wide standard. What changed in the last 24 hours is that the HASC FY27 chairman's mark is now live with markup June 4 — Congress is layering new AI mandates on top of unexecuted FY26 ones. (Contested — whether silence reflects a deliberate decision to subsume CFT functions under existing CDAO/USD(R&E) structure, or a genuine statutory compliance failure, is unknown; the January 2026 DoD AI Strategy favoring "speed and fewer constraints" supports the former reading, but no published directive confirms it.) Still uncertain: whether any HASC/SASC staff have drafted amendments targeting the CFT gap, and whether the FY27 AI subtitle substantively addresses the unbuilt FY26 infrastructure.

BIS rescinded the AI Diffusion Framework on May 13, 2025, committing to a future replacement. 381 days later: no Federal Register pre-publication notice, no ANPRM, no model-weight architecture decision. BIS has signaled it will skip long comment periods in favor of Interim Final Rules, compressing the window between signal and compliance obligation to 30 days or less. The decision implication for acquisition and bilateral negotiations: DoD cannot finalize allied AI technology-sharing terms with Five Eyes or QUAD partners until the replacement architecture is known. (Contested — export counsel disagree on whether the nonregulatory BIS guidance issued alongside the rescission is in practice more restrictive than the rescinded rule, because the "knowledge" standard can trigger licensing without a formal rule.) The HASC markup creates a parallel pressure: bills like H.R. 6996 (Full AI Stack Export Promotion Act) are already in play, and the longer Commerce waits, the more likely Congress writes the framework — potentially less flexibly than BIS would have.

CDAO is responsible for enabling the seven PSPs with monthly reporting to DepSecDef and USD(R&E), per the DoD AI Strategy. Swarm Forge (autonomous systems) and Agent Network (AI-enabled battle management) both carry DoDD 3000.09 obligations. The decision implication: the framing chosen for July demos — "experimentation" vs. "live operational capability" — will signal whether 3000.09 was applied or scoped out, and the latter creates congressional oversight exposure under NDAA 2025 Section 1061, whose waiver disclosure report (due December 31, 2025) remains unconfirmed as delivered. Still uncertain: whether any 3000.09 waivers have been sought, granted, or are being structured around demo scoping.

DOJ has charged a Google employee with using non-public information to trade on Polymarket, allegedly clearing ~$1M. The decision implication for ethics offices and program staff: federal personnel and contractors with access to predecisional AI program information, export-control timelines, or acquisition outcomes now sit closer to the insider-trading perimeter than generic tech employees. (Contested — some legal analysts argue existing insider-trading doctrine already covers prediction markets, but DOJ's theory in this case is untested; overbroad internal restrictions could chill legitimate red-team forecasting work.) No OGE, DoD Standards of Conduct, or agency GC guidance has issued.

[GOVERNANCE] CFT Standup Status: Whether CDAO has internally constituted the Section 1533 CFT via non-public directive, or whether June 1 will pass without action. Matters because: every AI vendor adjudication this week lacks an authoritative standard. Watch for: CDAO public notice, Federal Register filing, or oversight staff confirmation by June 1; silence past that date is a reportable statutory failure.

[EXPORT] BIS Replacement Rule Architecture: Whether the rule will control AI model weights, establish a trusted-partner bilateral framework, or revert to chip-count-only. Matters because: DoD's allied AI tech-sharing legal basis is undefined until known, and BIS's Interim Final Rule preference compresses lead time to ~30 days. Watch for: Any FR pre-publication notice, ANPRM, or State/NSC bilateral heads-of-agreement.

[INTEGRATION][AUTONOMY] PSP 3000.09 Compliance Posture: Whether Swarm Forge and Agent Network July demos have undergone 3000.09 review, and whether waivers exist. Matters because: undocumented autonomous demonstrations create both legal exposure and precedent for the broader PSP fast-follow pipeline. Watch for: DepSecDef or USD(R&E) Emil Michael framing of demos as "operational" vs. "experimentation."

[LEGISLATIVE] FY27 NDAA AI Subtitle Text: Whether the HASC chairman's mark substantively addresses the unbuilt FY26 CFT infrastructure or simply layers new mandates. Matters because: the seven-day amendment window closes June 4 and is the only near-term lever. Watch for: the chairman's mark PDF AI subtitle, and any staff-drafted amendments targeting the CFT gap.

Statutory foundations (FY26 NDAA Section 1533, BIS rescission press release, HASC mark release) are Tier 1 and assessed with high confidence. The CFT standup absence is assessed with moderate-to-high confidence based on absence of public signal at T-4 days — absence of evidence is not evidence of absence, but the silence is itself reportable. The Xiaomi MiMo pricing claim, the DOJ Polymarket profit figure, and the Anthropic/PrismML capability claims are vendor or single-source reports flagged accordingly. No breaking primary documents from the past 24 hours surfaced on CFT standup, BIS rulemaking, or PSP compliance posture — the silence at T-4 to a statutory deadline is the dominant signal of this issue, and decision-makers should treat the Known Unknowns as active collection requirements.

~1,380 words