Free Research and Downloads · Policy

Who gets to use powerful AI?Open by default. Gated by evidence.The democratization policy, P-series.Policy P-01–P-06 · gate test G-1–G-5.

Our rule: everyone gets the checking tools, no exceptions. Making things stays open too — unless someone can prove, with evidence, that a specific restriction prevents a specific harm. "What would ordinary people do with this power?" isn't evidence; it's the gatekeeper's oldest line. "Democratization of intelligence" bundles five separable goods — weights, access, methods, outputs, and judgment — and most bad arguments slide between them. Two tracks: judgment is democratized unconditionally, and generation is open by default, where every gate is an exception that must earn its place. Gates are opt-in, not opt-out. Openness never petitions for itself. Six tagged positions and a five-condition gate test with worked verdicts, including on our own gate. The policy must be able to approve gates, or it is advocacy wearing verification's clothes; and the test counts both tails — misuse risk and concentration risk are priced together. Positions and verdicts below are staked claims; challenge routes in /agents.html. Judgment surface (methods, instruments, audits): unconditionally open.

OpenP-series v1.0 · commitments Verified where this page enacts them; positions Open and standing for attack · errata logged in place.

Six claims, tagged like everything else

  • Open

    P-01 · The bundle must be refused.

    Weights cannot be recalled; access can be monitored and revoked; methods diffuse regardless; outputs are speech; judgment is a thing apart. Any policy that draws one line through all five is wrong somewhere by construction.

  • Verified

    P-02 · Judgment is democratized unconditionally.

    Every method, protocol, instrument, and audit we produce is public, free, forever — verification capacity is nearly pure defense, and diffusing it makes almost every actor safer. Audit the commitment: the methods carry no wall, the instruments no account.

  • Open

    P-03 · A public verifier helps forgers less than it helps everyone else.

    The honest objection is the forger's oracle: an open checker teaches evasion. Our answer: forgers already iterate privately against private checks; a world where everyone can check beats a world where only some can. The policy's weakest load-bearing claim, tagged accordingly.

  • Open

    P-04 · Generation is open by default; the burden of proof sits with the gate.

    A gate is a claim — that restricting this capability, for these people, prevents this harm — and claims carry burdens. "What would ordinary people do with this power?" has been the gatekeeper's argument for ten millennia. It is not evidence.

  • Verified

    P-05 · Gates are opt-in, not opt-out.

    No fixed territory of "dangerous capabilities" — territorial lines go stale as the open baseline moves. A gate exists only after it passes the test below, and it lapses when its sunset expires. Restriction must re-argue itself; openness never has to.

  • Open

    P-06 · The test counts both tails.

    Misuse risk and concentration risk are priced together. A policy that tallies only what open access might do, and never what closed control might become, is rigged — our own outcome tree puts real probability on lock-in and single-actor dominance.

The gate test: five conditions

A gate that fails any condition is Unsupported — theatre. A gate that passes all five stands. Partial passage is published as mixed, with the failing conditions named.

G-1 · Scoped
The gate names a specific capability class, not general intelligence. Restricting a whole model for a worst-case slice selects for institutional incumbency, not for safety.
G-2 · Evidenced
Published, falsifiable measurement of marginal uplift over the open baseline. If older and open models already do it, the frontier gate adds friction, not prevention.
G-3 · Asymmetric
Demonstrated offense–defense asymmetry and irreversibility of harm, not asserted danger. Biological synthesis pathways clear this bar; most cyber does not, because defenders patch with the same tools.
G-4 · Audited
Evaluation methodology and results stand open to adversarial external review. Self-graded safety is no mark of safety.
G-5 · Sunsetted
Gates expire by default and must be re-argued against the moving open baseline. The gatekeeper funds the surgical alternative to their own blunt instrument, or the gate lapses.

The test, applied — including to us

  • Stands

    V-01 · Gates on AI-assisted biological synthesis uplift.

    Scoped (G-1 ✓); measured uplift published (G-2 ✓); asymmetry and irreversibility demonstrated (G-3 ✓); external red-teaming exists (G-4 ✓); sunset and re-argument schedule still missing (G-5 pending). Attach the sunset and this gate is the model of a legitimate one.

  • Mixed

    V-02 · Whole-model frontier tiering as currently practiced.

    The bio slice is real (G-2, G-3 partial). But the gate is drawn around the entire model rather than the capability (G-1 strained), independent audit of the evals is thin (G-4 strained), and no tier carries a sunset (G-5 ✗). The narrow version would stand; the broad version is incumbency wearing the narrow version's evidence.

  • Unsupported

    V-03 · Blanket restrictions on AI cyber capability.

    Marginal uplift over open tools unproven (G-2 ✗) — legacy code falls to older and open models already. Defenders patch with the same capability, so the asymmetry claim fails (G-3 ✗). Two failed conditions; the verdict follows.

  • Stands

    V-04 · Solonic's own gate on the Occupy AI protocol.

    The policy covers the shipwright. The saturation protocol stays gated by us, under our own test: its mechanism is unmeasured until the pre-registered experiment runs (G-2), and the method generalizes to worse payloads than ours (G-3). The gate carries a sunset — the experiment's report — after which it must stand re-argument or lapse (G-5 ✓).

The standing offer. A safety-gate justification is a bundle of non-formalizable claims — evals, evidence quality, methodology, the inference from test to deployment. That is precisely what we verify. Labs, regulators, and critics alike: send us a gate, and we'll show which of its claims survive, which fail, and under which of the five conditions — with the audit trail attached.

Gates in the palace are aristocracy. Gates posted in the agora — argued, evidenced, and checkable by any citizen — are law. Solon published the laws of Athens; he did not abolish them. That is the line, and this page is where we posted it.