Beyond AGI Control: Sovereign Collective Intelligence as Complementary Infrastructure

Two of the most important voices on AGI risk — Stuart Russell and Leopold Aschenbrenner — agree on the diagnosis but differ on the cure.

Russell (Human Compatible, 2019): AI systems should defer to human preferences rather than pursue fixed objectives. Problem: whose preferences — and how are they authentically expressed?

Aschenbrenner (Situational Awareness, 2024): AGI arrives by 2027, security is catastrophically insufficient, state control becomes inevitable. Problem: this is control over AGI — but not sovereignty for citizens.

Both approaches share a structural gap: they assume preferences can be observed, aggregated, and managed top-down — by AI systems, corporations, or states.

BeTrueCore proposes a different layer entirely.

Rather than controlling AGI from above, it builds infrastructure through which humans continuously and verifiably express collective will — anonymously, cryptographically, with mathematically weighted intuition (Vote Weight Unit / VWU).

Three structural differences from existing approaches:

1. Energy: While AGI infrastructure demands gigawatts of centralised power — BeTrueCore processes biometrics locally (FaceID on device), generates ZK-proofs client-side, and archives proofs cheaply via Celestia DA. Sovereignty without the energy race.

2. Preference authenticity: MACI + ZK-SNARKs eliminate the preference falsification problem (Kuran, 1995) that undermines both Russell’s assistance game and democratic legitimacy claims.

3. Direction of control: Aschenbrenner’s «AGI realism» = top-down state control. BeTrueCore = bottom-up sovereign collective intelligence. AI as notary, not judge.

The question Russell asks — how do AI systems learn genuine human values? — cannot be answered by AI alone. It requires infrastructure where humans express those values freely, verifiably, and continuously.

That infrastructure does not yet exist.

BeTrueCore is a proposal for what it could look like.

Full specification: Zenodo | GitHub

Discussion welcome — especially on the canonical observation schema for VWU audit trails.