← Back to all writing

AI futures evidence — X5: US admin & elections

July 5, 2026

Futures index · 中文 · Main forecast

Each section: Claim · Why · Evidence · Analogue · Would update if · Conf (H/M/L).


Parent: crosscut x5 political administration ai
Extends: crosscut us governance capacity · node1 §culture-war · node6 · node2
Date: 2026-07-05
Format: Claim | Why | Evidence | Analogue | Would update if | Conf


TL;DR — top 5 P’s

PPoint est.Role
P(2028 admin materially shifts AI posture)0.50CX-ADMIN-FLIP — symmetric flip risk
P(BMIA delayed ≥24 mo | admin flip)0.55EV-ADMIN-28 timing sensitivity
P(GAAIA passes w/ dev preemption intact)0.12Scenario D tail; admin continuity modestly
P(AISI/CAISI adequately funded ≥$50M/yr by FY2028)0.35Institutional capacity for evals
P(no effective federal pause through 2028)0.88CX-NO-PAUSE — all X5 scenarios preserve

P = 0.50 — 2028-cycle administration materially shifts federal AI posture (CX-ADMIN-FLIP)

Claim: ~50% that the Jan 2029 administration implements ≥2 of 4 posture reversals (rescind 14365/14409 stack, restore AISI safety mandate, sign stronger audit statute, halt DOJ state-law challenges).

Why: Trump continuity is modal through 2028 but not certain; Dem coalition includes Encode/labor/natsec split; even R successor may shift (Hawley vetting vs Cruz preemption). Symmetric 0.50 = policy vector uncertainty, not raw party odds.

Evidence:

  • 05 crux registry — CX-ADMIN-FLIP Tier 3 queue
  • 08 parallel spines — S_soc admin row ◐
  • EO 14110 revoked 15 mo after signing — precedent for rapid reversal
  • research_ai_pause_advocacy_playbook.md §1.2 — executive fragility

Analogue: EPA WOTUS rules oscillate admin-to-admin; AI EOs follow same cycle faster.

Would update if: 2028 R nominee explicitly commits to 14365/14409 continuity + BMIA → revise to ≤0.35.

Conf: M


P = 0.38 — Scenario A: continuity acceleration (2026–2029 modal admin path)

Claim: ~38% weight: Trump-class posture persists through 2029 — EO stack operational, GAAIA dies or EO-substitute, states primary for frontier duties.

Why: Incumbent advantage; $43M+ AI super PAC alignment; industry preference for voluntary EO 14409; Jun 2026 same-month cluster suggests statute optional for moderate Rs.

Evidence:

Analogue: Dodd-Frank post-2010 — industry adapts to executive/regulatory frame when statute stalls.

Would update if: GAAIA signed with preemption before Dec 2026 → collapse A to ≤0.20.

Conf: M


P = 0.32 — Scenario B: divided government gridlock

Claim: ~32% that 2026–28 features split Congress or non-functional unified majority → S2 paralysis plus zero major AI signature by end-2028.

Why: Overlaps S2 P=0.55 governance paralysis but requires electoral split; 60-vote Senate threshold; Freedom Caucus + anti-preemption coalition repeatable; midterm stasis per NPR Jun 2026.

Evidence:

Analogue: 2011–2014 divided Congress — agency action substitutes for law.

Would update if: Bipartisan Commerce markup on BMIA and GAAIA before Nov 2026 → revise to ≤0.18.

Conf: M


P = 0.18 — Scenario C: reversal toward transparency + screening

Claim: ~18% that 2027–29 admin/Congress shift mandatory transparency (SB 53-class federal) + BMIA screening without dev preemption.

Why: Encode coalition + Dem Senate flip + EU AI Act pressure + screendna/Amodei lane; not full training pause. Requires electoral change or R fracture (Blackburn carve-out path).

Evidence:

  • GAAIA draft draws from SB 53/RAISE text (Trahan release, TechPolicy.Press Jun 2026)
  • node1 — P(SB53 compliance)=0.85
  • Nonce Media — EU floor forces US lab compliance infrastructure Aug 2026

Analogue: Post-2008 financial reform — crisis + coalition shift after delay.

Would update if: Dem wins 2028 and first 100 days include BMIA signing → revise up to ≥0.30 for scenario weight.

Conf: M


P = 0.12 — Scenario D: reversal toward preemption-only (weak federal floor)

Claim: ~12% GAAIA or successor passes with 3-yr dev preemption + weak audit enforcement — industry capture tail.

Why: Same as Node 6 O1 P=0.12 but conditional on admin continuity + industry lobbying ($8.5M Q1 2026 preemption); House Ds opposed Jun 2026 draft but R unified govt could force floor.

Evidence:

  • node6 — P(GAAIA preemption intact)=0.12
  • McDonald Hopkins / ArentFox — GAAIA preemption is “most controversial”
  • Playbook §2.1 — preemption = highest negative lever

Analogue: Telecom Act 1996 preemption of local entry barriers.

Would update if: Preemption stripped in committee with R chair → revise to ≤0.05.

Conf: M


P = 0.55 — BMIA delayed ≥24 mo conditional on EV-ADMIN-28 admin flip

Claim: If Jan 2029 admin flip occurs, ~55% BMIA slips ≥24 mo vs baseline (2027-12 target missed → 2029-H2+).

Why: New appointees 6–12 mo; Commerce committee restart; 119th session momentum lost; physical-layer bills eventually pass but admin transitions add full Congress cycle. Not conditional on flip direction — either party reset causes delay.

Evidence:

  • BMIA referred Jan 29 2026, no floor vote Jul 2026 — already 6 mo into delay
  • crosscut us governance capacity — BMIA ×0.70 under S2 → 0.32 by 2027-12
  • COVID vaccine EUA — admin transitions delayed non-emergency statute ~12–18 mo

Analogue: ACA implementation delay 2017 repeal fight — regime uncertainty stalls narrow bills.

Would update if: BMIA signed Dec 2026 → conditional irrelevant; revise base BMIA P up.

Conf: M


P = 0.45 / 0.60 — BMIA by 2027-12 / 2028-06 (X5-adjusted baseline)

Claim: Baseline BMIA passage unchanged from Node 2 at S0; apply X5 scenario multipliers from parent modifier table.

Why: BMIA is admin-insensitive relative to GAAIA — Cotton–Klobuchar bipartisan; screendna Jun 2026; physical chokepoint not ideology-coded as pause.

Evidence:

Analogue: See Node 2 BMIA sections.

Would update if: BMIA dies committee without substitute → ≤0.25 / ≤0.40.

Conf: M


P = 0.12 — GAAIA passes with 3-year state dev preemption intact (admin continuity ↑)

Claim: ~12% — same Node 6 O1; under Scenario A/D conditional P(preemption intact | GAAIA floor)0.55.

Why: Admin continuity raises industry confidence to force floor vote; still blocked by Encode + states + 99–1 anti-preemption precedent on different frame.

Evidence:

  • node6 — O1 0.12; P(preemption stripped | floor)=0.55
  • Roll Call Jun 2026 — bipartisan draft, preemption fight
  • Wiener/Encode opposition (playbook §313)

Analogue: See Node 6 O1.

Would update if: White House endorses preemption text in signing statement draft.

Conf: M


P = 0.18 — GAAIA passes, preemption stripped (Scenario C path)

Claim: ~18% federal audit law without gutting SB 53/RAISE dev obligations — compromise path.

Why: Obernolte explicitly solicited civil-society feedback; SB 1047→SB 53 negotiation template; Dem commission opposition to preemption not to audits.

Evidence:

  • node6 — O2 0.18
  • TechPolicy.Press — Trahan/Obernolte “not one framework to rule them all”
  • Anthropic endorsed SB 53, criticized caps

Analogue: See Node 6 O2.

Would update if: R hold + industry kills bill entirely → revise to ≤0.10.

Conf: M


P = 0.35 — AISI/CAISI adequately funded (≥$50M/yr + ≥100 FTE by FY2028)

Claim: ~35% NIST AI center reaches adequate capacity for frontier evals (not just DeepSeek one-offs).

Why: H.R. 9363 proposes $20M/yr; GAAIA $100M/yr — both not enacted; NIST chronic underfunding; shutdown/RIF precedent; Kratsios innovation framing safety-scale budget.

Evidence:

Analogue: CISA pre-2015 — statutory mandate years before adequate appropriations.

Would update if: H.R. 9363 signed with ≥$50M → revise to ≥0.55.

Conf: M


P = 0.65 — CAISI remains innovation-framed (not full AISI restoration) through 2028

Claim: ~65% Trump CAISI mission persists through 2028 even if Dem wins — full AISI safety restoration delayed to statute.

Why: Rebrand survived 12 mo; Lutnick/Kratsios public framing; RIF/agency culture; Dem platform may accept “security and innovation” hybrid (Banks letter lane).

Evidence:

  • Jun 2025 CAISI rebrand (Commerce)
  • node6 Phase 2b — AISI→CAISI modal dual-track
  • Banks Jun 2026 — CAISI visibility, not NSA solo

Analogue: DHS post-9/11 — mission creep to security without name revert.

Would update if: Dem 2029 Day 1 AISI re-rebrand → revise to ≤0.40.

Conf: M


P = 0.88 — No effective federal training pause through 2028 (CX-NO-PAUSE, X5-stable)

Claim: All four X5 scenarios preserve ~88% P(no pause) — admin flip moves vetting/screening, not training moratorium.

Why: Culture-war equilibrium (Node 1 P=0.62 vetting modal); Cruz 99–1; natsec anti-pause; Anthropic conditional expect not commit; Scenario C transparency not halt.

Evidence:

  • 05 crux registry — CX-NO-PAUSE 0.88
  • node1 — P(federal training cap)<0.05
  • Playbook §1.5 — conditional pause mainstream, not full stop

Analogue: See CX-NO-PAUSE node refs.

Would update if: Binding training moratorium signed → falsify crux.

Conf: H


P = 0.55 — US governance paralysis S2 through 2028 (CX-US-PARALYSIS, X5-B overlap)

Claim: Scenario B embeds S2; other scenarios reduce paralysis probability — net ~55% unchanged from S2 crux when weighted.

Why: X5 does not replace S2 — adds electoral mechanism. Continuity (A) may pass BMIA despite shutdown history.

Evidence:

Analogue: See S2.

Would update if: GAAIA or BMIA signed before Dec 2026 → S2 and CX-US-PARALYSIS revise ≤0.35.

Conf: M


P = 0.40 — CX-BMIA holds (BMIA / mandatory screening by 2027-12)

Claim: X5-adjusted ~40% — slight from Node 2 0.45 accounting for admin-flip delay mass (0.50 × 0.55 delay).

Why: Weighted average across scenarios; EU/IGSC offset partially.

Evidence:

  • node2 §P=0.45
  • X5 parent modifier table

Analogue: See CX-BMIA / Node 2.

Would update if: BMIA signed → ≥0.85 conditional on enforcement.

Conf: M


P = 0.70 — 2026 midterms change ≥1 load-bearing committee chair (AI jurisdiction)

Claim: ~70% House E&C or Senate Commerce chair changes party or person Nov 2026 — normal midterm turnover + competitive map.

Why: Historical midterm seat swing; AI jurisdiction concentrated in few chairs (Obernolte, Cruz-class); even held chamber often rotates subcommittee gavels.

Evidence:

  • NPR Jun 2026 — election shapes 120th Congress AI agenda
  • Legis1 Jun 2026 — 10 AI bills advanced under current chairs

Analogue: 2018 Energy & Commerce leadership change stalled tech bills 6 mo.

Would update if: R expand majority with same chairs retained → revise to ≤0.45.

Conf: L–M


P = 0.25 — R unified government functional for AI statute 2027–28

Claim: ~25% R holds White House + both chambers and passes any GAAIA/BMIA-class bill before end-2028.

Why: Internal R split (preemption vs carve-outs); 60-vote Senate; Freedom Caucus CR tactics; BMIA more likely than GAAIA alone (~0.45).

Evidence:

Analogue: 2017 R trifecta — only TCJA cleared filibuster bar for major item.

Would update if: BMIA floor vote scheduled Q4 2026.

Conf: M


P = 0.30 — D Senate flip in 2026 midterms

Claim: ~30% Democrats gain Senate majority Nov 2026 (speculative — Jul 2026 forecast band).

Why: Standard competitive midterm map with incumbent R exposure; AI not sole driver but jobs/affordability narrative helps Ds; super PAC counterweights.

Evidence:

  • GZERO / Time 2026 midterm AI salience
  • OpenSecrets AI PAC spend both sides

Analogue: 2006 midterm Senate flip — war economy issue replaced by local economics; AI may play similar 2026.

Would update if: R gain seats Nov 2026 → revise to ≤0.10.

Conf: L (election forecast)


P = 0.85 — SB 53 / RAISE enforce on schedule regardless of X5 scenario

Claim: State frontier laws effective Jan 2027 with ≥85% enforcement continuity — federal admin flip does not repeal state law.

Why: SB 53 signed; RAISE amended Mar 2026 to track SB 53; federal litigation may delay but not void; Scenario D preemption is only material downgrade.

Evidence:

  • node1 — P(SB53)=0.85
  • Carnegie Feb 2026 — state enforcement capacity critique (separate from existence)
  • Law.com Jan 2026 — EO 14365 vs Jan 1 effective dates unsettled

Analogue: CA privacy (CCPA) enforced despite federal absence.

Would update if: Supreme Court blocks SB 53 dev duties → revise to ≤0.50.

Conf: M–H


P = 0.45 — DOJ AI Litigation Task Force files challenge to blue-state law by end-2027

Claim: ~45% federal lawsuit vs CA/NY/IL frontier law before 2028 election — political timing favors 2027 filing.

Why: EO 14365 mandates Task Force; Jan 2026 DOJ memo; no filing Jul 2026 — delay may be strategy; Commerce onerous list still pending.

Evidence:

Analogue: DOJ vs CA net neutrality — multiyear litigation without immediate strike-down.

Would update if: Task Force disbanded Jan 2029 → path dead.

Conf: L–M


P = 0.60 — EO 14409 voluntary framework operational by Q1 2027

Claim: ~60% NSA-led “covered frontier model” designation + voluntary 30-day review functioning for ≥2 frontier labs.

Why: 60-day deadline from Jun 2026 → Aug 2026 design; industry incentive to participate (gov cyber clearinghouse); not legally mandatory — OpenAI/xAI may opt in partially.

Evidence:

  • EO 14409 §3 — 60-day framework design
  • Perkins Coie Jun 2026 — voluntary emphasis
  • node6 — EO-only 0.25 as substitute for GAAIA

Analogue: Vulnerability Equities Process — voluntary gov access norm.

Would update if: Zero lab participation announced by Mar 2027 → revise to ≤0.25.

Conf: M


P = 0.72 — EU AI Act sets de facto compliance floor for US frontier labs (Aug 2026+)

Claim: ~72% Anthropic/OpenAI/Google-class labs maintain EU-grade GPAI documentation for global products regardless of X5 US admin scenario.

Why: Aug 2 2026 applicability; extraterritorial scope; product economics favor single compliance stack; Nonce Media “compliance asymmetry” argument.

Evidence:

Analogue: GDPR — US companies complied for global ops without US statute.

Would update if: Major US lab exits EU market → revise to ≤0.40.

Conf: M–H


P = 0.28 — EU accelerates US federal legislation via competitive pressure (Brussels Effect on Congress)

Claim: ~28% EU enforcement before US statute increases Congress urgency to pass GAAIA/BMIA-class bill 2027–28.

Why: Brussels Effect on companies is high (0.72 above); on Congress is weaker — US federal vacuum persists by design under Trump; EU Digital Omnibus may slow EU own deadlines.

Evidence:

  • Trilligent 2025 — “end of Brussels Effect?” under US deregulation
  • Internet Policy Review — experimentalist governance vs Brussels Effect for AI Act
  • Nonce Media — “watched closely in Washington”

Analogue: GDPR did not produce US federal privacy law by 2026.

Would update if: GAAIA markup explicitly cites EU Aug 2026 deadline → revise to ≥0.40.

Conf: L–M


P = 0.62 — Culture-war vetting coalition modal wins (N1 ext, admin-stable)

Claim: ~62% BMIA + SB 53 + GUARD-class vetting advance; no federal training pause — stable across X5 except Scenario D tail.

Why: See Node 1 culture-war section; admin changes reweight not replace.

Evidence:

Analogue: See Node 1.

Would update if: Federal training moratorium signed.

Conf: M


P = 0.20 — Anti-regulation fusion (GAAIA O1 tail) under continuity admin

Claim: ~20% broad preemption without ELVIS/KOSA/NO FAKES carve-outs — Node 1 tail; under Scenario A/D.

Why: $8.5M preemption lobbying; Trump EO stack; industry wants uniformity.

Evidence:

  • node1 §P=0.20
  • Playbook §2.1

Analogue: See Node 1 anti-regulation tail.

Would update if: Blackburn carve-outs explicitly in GAAIA final text → ≤0.08.

Conf: M


P = 0.15 — Biden-era 14110-style compute reporting restored by 2029

Claim: ~15% mandatory compute/training reporting above threshold returns even under Scenario C — full 14110 restoration unlikely (Beat China frame).

Why: 14110 revoked completely; Dems may prefer SB 53 federalization over compute caps; hardware metering still immature.

Evidence:

  • Longterm Wiki E366 — 85% completed then revoked
  • Playbook §1.5 — verification window closing
  • SB 1047 veto — caps dead in CA

Analogue: Assault weapons ban restoration attempts — symbolic, rarely full revert.

Would update if: GAAIA includes training registration title → revise to ≥0.30.

Conf: L–M


P = 0.08 — Populist moratorium floor vote (Cruz-class, admin-sensitive)

Claim: ~8% R-led moratorium floor vote — slightly under continuity if used as preemption negotiation lever.

Why: Cruz BBB stripped 99–1; still in Overton window; not admin priority.

Evidence:

  • node1 §P=0.08
  • Playbook §1.2 Cruz dynamic

Analogue: See Node 1.

Conf: L


P = 0.42 — House advances CAISI codification (H.R. 9363 or successor) by end-2027

Claim: ~42% statutorily establishes NIST AI center — may be underfunded ($20M).

Why: Bipartisan Science Committee interest; Obernolte/Kratsios alignment; same as adequate funding P=0.35.

Evidence:

Analogue: NIST cybersecurity framework codification — years from bill to budget.

Would update if: H.R. 9363 dies in appropriations → ≤0.20.

Conf: M


P = 0.33 — S.3952 (Future of AI Innovation Act) passes Senate before 2028

Claim: ~33% Young–Cantwell–Blackburn innovation/standards bill becomes law — parallel track to GAAIA, less preemption fight.

Why: Bipartisan cosponsors Feb 2026; star print Mar 2026; may attach to NDAA or lame duck.

Evidence:

Analogue: CHIPS Act — standards/innovation can pass when omnibus safety stalls.

Would update if: Bill dies committee Dec 2027 → ≤0.15.

Conf: M


Claim: ~3%same as S2; admin change crux does not materially move this tail.

Why: See crosscut us governance capacity §P=0.03 — link only.

Evidence: S2 civil war section; Brookings shutdown analysis.

Analogue: See S2.

Would update if: See S2 falsifiers.

Conf: L


Actor table — Congress, White House, labs, EU

ActorAdmin alignment (2026)Key leverP(modal action)X5 sensitivityConf
Trump White HouseAccelerationEO 14365/14409; FrameworkP(EO wins over GAAIA)=0.25High — defines Scenario AM
DOJ AI Litigation Task ForceAccelerationSue blue statesP(filing by 2027)=0.45High — flip ends active casesM
House E&C (Obernolte)Bipartisan safety + R innovationGAAIA markupP(markup 2027)=0.50Chair flip → preemption M
Senate Commerce (Cruz)R uniformityBMIA path, preemptionP(BMIA report)=0.55Natsec framing survives flipM
Senate Judiciary (Hawley)Populist vettingGUARDP(GUARD advances)=0.70Low admin sensitivityM
Blackburn (R-TN)Selective preemptionCarve-outs gatekeeperP(carve-outs hold)=0.65Med — blocks Scenario D tailM
Encode / WienerState transparencyAnti-preemptionP(SB53 stands)=0.85High — Scenario C allyH
OpenAIIndustry pragmatistVoluntary EO + GAAIA negotiateP(opposes dev preemption)=0.55MedM
AnthropicSafety + EU compliancePublic First PAC; SB 53 endorseP(EU-floor compliance)=0.90Low — admin-independentH
xAI / MuskAccelerationAnti-”woke AI” narrativeP(jobs frame wins)=0.85High — Scenario A mascotM
a16z / Big TechPreemption$8.5M+ lobbyingP(federal cap)<0.05MedH
AFL-CIO Workers FirstLabor bridgeDisplacement, not pauseP(sustains agenda)=0.75Med — Scenario C allyM
EU AI OfficeExternal floorGPAI Aug 2026 enforcementP(de facto US lab compliance)=0.72Low — orthogonal to US adminM–H
Leading the Future PACPro-innovation RMidterm spendP(continuity R)=0.55HighL–M
Public First Action PACSafety oversight D/RMidterm spendP(Senate safety caucus)=0.35MedL–M

Crux modifier summary — CX-BMIA, CX-US-PARALYSIS, CX-NO-PAUSE

CruxBaseline PX5 modifier (how to apply)
CX-BMIA0.40Scenario A ×1.05; B ×0.70–0.75 (S2 stack); C ×1.20–1.25; admin flip delay +12–24 mo mass
CX-US-PARALYSIS0.55Scenario B = primary trigger; A ×0.85; C with unified D ×0.70
CX-NO-PAUSE0.88Scenario A/D ×1.02–1.04; C ×0.96–0.97 (vetting ↑, pause still unlikely) — ≤3pp relative

Do not multiply CX-US-PARALYSIS × Scenario B × S2 elevated paralysis as independent — use parent stitching rule.


p(doom) direction (X5 evidence summary)

X5 pathBio tailMisalignNo pauseWhimper
A continuityModalModal raceModal
B gridlock +3–6pp cond. T2Ambiguous↑↑
C transparency BMIA ↑±0–1pp↓ modest↓ modest
D preemption weak screening 1–2pp↑↑

Update log

DateChange
2026-07-05Initial 28 P sections + actor table + crux modifiers