Structural Defenses Against AI-enabled Extreme Concentration of Power

Raising $22,720active

Decentralising government and private data infrastructure through next-generation distributed computing architectures to protect information sovereignty and prevent AI-enabled concentration of data as power.

Project Details

We will build and drive adoption of Grove, a one-stop collaborative workspace for sovereign data. Grove provides identity and access control, file storage, chat, calendars, contacts, and collaborative editing (the everyday tools hyperscalers provide) at market-competitive quality. But Grove is the first of its kind: a collaborative platform built on RAIN, our state-of-the-art, proven fault-tolerant replication system, currently being rolled out to production as the storage orchestrator for hundreds of European companies. RAIN lets any online application run across independent servers, with no single provider able to disrupt, censor, or extract data. The work is delivered by a three-party consortium: RAIN Sàrl develops Grove; the Reliable Information Laboratory leads the underlying science, dissemination, and Swiss-administration engagement; and JACONA PBC develops US contracting and distribution.

Theory of Impact

Information is power. Centralised information concentrates power. AI makes abusing concentrated power easier. Decentralising information systems is therefore a direct structural defence against AI-enabled abuses of power. Currently, a handful of chokepoint hyperscalers hold the working knowledge of most democratic institutions: a single executive order, compromised vendor, or act of coercion can expose voter rolls, freeze allied states' operations, or hand an adversary a decisive intelligence advantage. RAIN (and applications built on it, like Grove) dissolve this chokepoint: no government or private actor can quietly tap data it has no right to see or quietly suppress data it must release. On RAIN, every institution can run its own Grove instance (and any other applications) with two orders of magnitude better availability than managed cloud services (experiments show less than three seconds of downtime per year on commodity hardware), cryptographic guarantees against tampering, and no vendor able to read or revoke content. Specifically, this enables much needed transparency and accountability for AI use in government (i.e. allowing the US Congress to audit how is the US executive branch is implementing and using AI) as well as structural check against mass surveillance.

People

Briant Jacobs

Team Member

Grants Received– no grants recorded

Add grant

Funding Asks

grantmaking.ai Launch Round

Applied

Minimum

$22,720

Ideal

$42,450

How the money will be spent

This grant will fund the business side cost to bring the public benefits of RAIN to the world. This includes a professional workspace and internal systems (Proton Workspace Premium), team accounts on Claude Max (6 subscriptions), domain registration and hosting fees for the consortium, RAIN Sarl and JACONA PBC, one additional patent application filing, copyright/trademark legal consultation fees.

The core system is built and running, not on paper. RAIN is about 40,000 lines of Rust, with all planned features implemented and 388 tests passing. We have had it live in a real 6-node European deployment, where it held up at over seven-nines uptime (99.99999%) across more than 15 million seconds of operation. So we have real production data, not just benchmarks. Three patents have been filed on this work, and all the IP is assigned, so there are no outside dependencies. The products built on top are underway too: Grove is in active development, with the frontend stack chosen and collaborative editing as the next milestone. What is left is turning a working platform into products individuals and organizations can implement and integrate.

Discussion

No comments yet. Be the first to share your thoughts.