Project summary

Generative AI has made fake and manipulated media cheap and everywhere — and most tools that check authenticity ask you to trust a platform or an AI model. IdentityTrusts flips that: an open primitive that lets anyone verify content-and-consent claims on their own machine, offline, using open mathematics with no trusted model in the verification path. A working version exists today; this funds turning it into honest, documented public infrastructure anyone can run and build on.

What are this project's goals? How will you achieve them?

IdentityTrusts is a provenance primitive built so the recipient of a piece of media can verify its provenance and a bound consent record entirely on their own machine — offline, with open tools, indefinitely — using no trusted model in the verification path and no service to call. The trust basis is standard public-key signatures and hashing: open mathematics anyone can inspect, not a black box you have to believe.

Two properties make it more than a watermark:

• Sovereign verification. A recipient confirms a claim without contacting a server, trusting a platform’s say-so, or running code they can’t read. The claim survives the disappearance or compromise of whoever issued it.

• Unlinkable by design. Each consent record carries a per-record nonce, not a stable identifier — provenance is verifiable without turning the system into tracking infrastructure. Consent without surveillance.

How will this funding be used?

Modest independent-developer rate; software/spec/docs only, no hardware. Two tiers:

At minimum funding ($9,000) — the safety-relevant core:

• Honest robustness characterization on realistic data. Characterize the primitive’s behaviour across diverse, realistic media and publish an honest envelope: which transform categories it survives and which it does not (including that cropping is not currently recoverable), plus the methodology. Following responsible-disclosure norms, the published envelope tells adopters whether the primitive fits their use case without shipping a turnkey evasion recipe alongside the open code. This is the work that turns the primitive from demonstrable into something people can actually rely on within stated bounds.

At full funding ($26,000) — also:

• Standalone client-side verifier (open, browser-based + CLI) anyone can run with no service dependency and no trusted weights, embeddable in exported proof packages.

• Open specification of the consent-record structure and verification procedure, precise enough that a third party can build an interoperating verifier from it alone.

• Developer documentation and a plain-language explainer of the sovereignty and privacy properties, so non-cryptographers can adopt it.

Who is on your team? What's your track record on similar projects?

Solo independent builder. Paul Krause out of San Diego, California. Relevant track record:

• Advised multiple companies in the AI, cybersecurity and Cryptography space from 2016 to present.

• Founded and shipped a project that reached real scale and public visibility (10,000-item collectible drop sold out in 12 days, ~$20M valuation, Forbes coverage). (Include only if you choose to attach your real identity — it’s a strong credibility signal here, but it’s your OPSEC call.)

• Built production multi-agent orchestration and DeFi/on-chain tooling (autonomous strategy agents, DEX integration), with cryptographic signing, append-only signed-event logging, and Bitcoin-anchored timestamping already in production use.

• Built the working IdentityTrusts provenance primitive being funded here — embed/recover/verify, classical and fully client-side, demonstrable today.

The work is real, it ships, and the verification trust basis is open and model-free by design — which is both the technical character of the project and its ethical commitment.

What are the most likely causes and outcomes if this project fails?

Honest failure modes:

1. The robustness envelope turns out narrower than hoped. Realistic-data characterization may show the primitive degrades under more transforms than expected. Outcome: still a valuable, publishable honest negative result that tells the field where this class of construction holds and where it doesn’t — and the verifier and spec still ship as public goods. Honesty about limits is part of the deliverable, not a failure of it.

2. Adoption doesn’t materialize quickly. Provenance infrastructure adoption is slow. Outcome: the open primitive, spec, and verifier remain permanently available for others to build on; nothing is lost to the commons.

3. Builder capacity. As a solo builder I’m the bottleneck. Outcome: I scope deliverables so the minimum-funding core (the characterization) is independently valuable and completable even if the larger scope slips. The tiered ask reflects this — the minimum tier is a complete, shippable unit on its own.

What does not fail safely is overclaiming, so I don’t: every claim here stays inside what’s been measured, and the characterization milestone exists specifically to replace optimism with honest, documented bounds.

How much money have you raised in the last 12 months, and from where?

Full transparency: $0 raised to date for this work; it has been self-funded.

One pending application to disclose: an application to the NLnet/NGI Zero Commons Fund (EU) is in review for a broader scope of the same open-source work (€40,200, submitted late May 2026, outcome expected mid-to-late 2026). If both this and NGI are funded, I will not bill the same milestone to both — Manifund would fund the faster near-term core (and/or complementary scope) while NGI funds the broader program, with the split documented openly. I’d rather disclose the overlap up front than have a funder discover it.