Information flow ↓ · ai artefact
Closed weights mandate
Open weights are irrecoverable once released and any proliferated model becomes misuse infrastructure; state classification is the only reliable non-proliferation regime.
Mechanism
Classify frontier weights, training code, and capability evaluations as national security information with Espionage Act penalties for release.
Falsification signal
Classified capability appears unclassified within two years.
A strategy held without a falsification signal is not strategy; it is affiliation. Continued support after this signal lands is identity, not bet. See the identity diagnostic.
Self-undermining threshold
overshoot riskWhen state custody exceeds state competence to secure.
Concentrates frontier capability inside the state apparatus most capable of losing it. The nuclear-secrets precedent is the reference case.
Every strategy has a stable region where it reinforces itself and an unstable region where pursuit defeats it. The threshold between them is usually narrower than advocates acknowledge.
Historical analogue
Nuclear · Atomic Energy Act classificationEvery strategy inherits a plausible ceiling from its precedent. The analogue conditions the realistic reach.
Produced
Concentrated control of nuclear secrets inside the state apparatus.
Did not produce
Secrets leaked across borders anyway (Rosenberg, Pakistan, Iran, North Korea).
Addresses 2 failure scenarios
all scenarios →People on the record
2Profiled figures appear first, with their tier in small caps. Each face links to the person and their full quote record. Tag: closed-weights.
Jeffrey Ladish
Applied or adjacent technical · Recognised inside subfield
Lawrence Lessig
Harvard Law professor; Creative Commons founder
Coordinates
Conflicts, grouped by mechanism
3Lever opposition
same lever, opposite pullThe pair's primary lever is the same; they pull it in opposite directions. A portfolio containing both is internally incoherent on that lever.
Frame opposition
incompatible premisesThe strategies accept different premises about what AI is or what the binding problem is. They conflict not on lever choice but on the frame that makes lever choice sensible.
Complements, grouped by mechanism
4Same phase, different layer
same stage, distinct leversBoth are active in the same phase of the transition but act on different layers (model vs institution vs culture). They cover different failure modes inside the same window.
Stage-sequenced
one sets up the otherThe pair is phase-offset: one acts before the transition, the other during or after. The first creates the conditions under which the second binds.
Axis position
Source note: Closed weights mandate strategy.md