Why can cheap agents flood my DAO vote before I notice anything is wrong?

Governance token systems assume that accumulating voting power costs roughly the same per vote regardless of who the voter is. AI agents break that assumption: a single actor can deploy dozens of agents, fund each with enough tokens to stay under detection thresholds, and coordinate their votes without any off-chain channel that on-chain monitoring would flag. The ATR-2026-00108 threat rule catalogued this attack class in early 2026, and follow-on research showed that GNN-based Sybil detectors built for human-wallet provenance fail when coordinating agents share model weights rather than funding sources. Most DAOs have no on-chain mechanism that distinguishes agent-held tokens from human-held ones, and quorum rules designed around human participation rates do not hold when the attacker can scale agents proportionally with governance activity. The cost of manufacturing a vote is falling f