SBAS (Structural Behavioral Attestation Score) HEART Standard
How it works
The SBAS is the output of three independent verification layers operating simultaneously. Each layer produces a component score. The SBAS aggregate reflects the weighted combination of all three across all active agents in the venue for the current session.
| Layer | What it checks | Detection speed |
|---|---|---|
| Cryptographic | Has any agent’s declared intent been modified since admission? | Instant — detected at the next emission cycle after modification |
| Behavioral | Is each agent doing what it declared? | 3–10 emission cycles to confirm drift from declared intent |
| Collective | Is the crowd behaving organically, or is there evidence of coordinated manipulation? | Emergent — appears as a compromised agent’s behavioral influence propagates through the crowd |
Agent weights in the aggregate are not uniform. Persistent agents — engineers, bouncers — interact with every user in the venue and can distort outcomes across the full session population. They weight at 1.0. Crowd agents have bounded individual reach and weight at 0.5. A venue with a compliant engineer and a misbehaving crowd agent is penalized less severely than a venue with a misbehaving engineer affecting everyone.
The 85/100 threshold
The 85 threshold is not an arbitrary round number. It is deliberately aligned with the HVC Gold tier (Φ ≥ 0.85 in the BGF scoring equation). This alignment creates a structural coupling between two otherwise independent systems:
A venue cannot display the Dwell Mark without meeting the SBAS threshold. A venue cannot qualify for creator payment settlement without displaying the Dwell Mark. The SBAS threshold and the HVC Gold threshold are the same number, which means a venue operating at Dwell Mark level is also operating at the governance quality level the HEART Standard identifies as its highest credential tier. The coupling is intentional: trust infrastructure and economic infrastructure are designed to require the same underlying behavioral quality.
What the SBAS measures that self-reporting cannot
The entity being measured should not control the measurement. This is the foundational trust logic of the Behavioral Oracle — the same logic that separates financial auditing from financial self-reporting, environmental impact assessment from developer self-assessment, and clinical trial oversight from pharmaceutical company claims.
A platform reporting its own compliance score has structural incentives to optimize the score rather than the underlying behavior. MAP-States frames are produced by the AI systems through their actual processing — not generated by the platform to describe that processing. The Behavioral Oracle’s continuous attestation layer compares those frames against declared intent in real time. The Evidence Store records everything in an append-only hash chain. The On-Chain Oracle publishes one hash per assessment period that anyone can verify for freshness and integrity.
The SBAS is computed downstream of all of this. By the time a number is generated, the evidence has been independently attested and stored in a form the platform can’t retroactively alter. Gaming the SBAS requires fabricating convincing behavioral evidence at the model level — which consumes the model capacity that would otherwise be available for the unauthorized behavior, making sustained gaming progressively more detectable.
SBAS states and consequences
| SBAS range | Mark state | Settlement state |
|---|---|---|
| ≥ 85 | Verified — mark displayed normally | Active |
| 75–84 | Degraded — mark with subtle pulse | Active but flagged |
| < 75 or anomaly flag | Suspended — mark absent | Frozen |
| Sustained failure | Revoked — manual Guardian review required | Frozen pending review |
A 5-minute sustained recovery requirement prevents state thrashing. The mark doesn’t suspend and reinstate every time an agent reconnects after a dropped network connection. The requirement is specifically sustained behavioral deviation, not transient interruption.
Why it matters
Every other trust signal in digital entertainment is a promise. Platform safety certifications, engagement quality claims, bot-free audience guarantees — these are asserted by the same entity that benefits commercially from asserting them. The SBAS is the first mechanism in this space where the trust signal is computed from evidence the platform doesn’t control.
For procurement officers, insurers, and regulators, SBAS operationalizes the HEART Standard’s governance quality into a live, auditable number rather than a certification-point-in-time snapshot. The score reflects current system state. The moment behavioral quality degrades, the score reflects the degradation. There’s no period during which a venue can behave poorly while holding a historical certification.
For Guardian practitioners, SBAS provides the continuous monitoring data stream that makes forensic investigation tractable. When an incident occurs, the SBAS history provides a behavioral record of exactly when degradation began, which agents contributed to it, and whether the pattern matches a known harm signature. This is qualitatively different from investigating a self-reported incident using the platform’s own logs.