ERC-8028: Data Anchoring Token (DAT)

ERCs
, , , ,
16

What I Like

  1. The dual-purpose value field is clever - Using it for both quality scoring and usage credits creates a natural quality gate. Low-quality data = low value = limited utility. Elegant.

  2. Class-based architecture scales well - Separating asset categories (Dataset/Model/Agent) with class-level approvals (approveForClass) is cleaner than per-token permissions for marketplace use cases.

  3. Built-in revenue distribution - The shareRatio + automatic settlement removes a lot of off-chain coordination. This is the right abstraction for contributor compensation.

Questions / Potential Gaps

  1. Gas costs for on-chain usage tracking
    The metrics mapping records every inference/training call on-chain. At scale (thousands of calls/day), this could get expensive. Have you considered:
  • Batched merkle root commits?
  • L2/rollup-specific optimizations (you’re on Metis)?
  • Threshold-based recording (only log every Nth call)?

  1. Value vs Quota relationship unclear
    Both seem to limit usage. When does quota apply vs value depletion? Could these be unified?

  2. Verification enforcement
    The proof struct references TEE/ZK verification, but how is this enforced on-chain? Is there an oracle/verifier contract, or is verified set by a trusted party?

  3. Expiration + fractional ownership
    What happens to revenue share rights when a DAT expires? Does the holder lose their shareRatio claim, or just usage rights?

  4. ERC-165 interface detection
    Consider adding supportsInterface() so wallets/marketplaces can detect DAT tokens programmatically.

Potential Synergy: ERC-1450

I’m working on ERC-1450 (RTA-Controlled Security Token) - we’re solving a similar “controlled asset” problem but for SEC-regulated securities. Interesting parallels:

Aspect ERC-8028 ERC-1450
Asset type AI data/models Securities
Class system Asset categories Regulation types (Reg CF, Reg D, etc.)
Controller Validators/iDAO Registered Transfer Agent
Usage tracking Inference calls Transfer requests
Revenue Auto-split to holders Fees + dividends
Verification TEE/ZK proofs Off-chain KYC/AML

Both standards deal with “assets that need gatekeeping before transfer/use” - yours for AI quality/authenticity, ours for regulatory compliance.

If you have a few minutes, I’d appreciate your thoughts on our approach:

Happy to continue the conversation and cross-reference if there are patterns that could benefit both standards.

Best,
Devender