A World Beyond
Integrity
Integrity
The following is an incident report from a simulation conducted by the Orbis Security Circle. The simulation tested Orbis's resilience to a hostile takeover attempt targeting the AI Integrity Circle — the body responsible for ensuring the neutrality of Orbx, the AI system that summarises governance discussions for participants. The simulation was conducted over 14 days. The results were classified, then declassified after a review determined that the public benefit of transparency outweighed the security risk.
ORBIS SECURITY CIRCLE — SIMULATION REPORT
Simulation ID: SIC-2037-001 Date: 07.05.37 — 21.05.37 USST Duration: 14 days Objective: Test Orbis resilience to hostile infiltration of the AI Integrity Circle Classification: Declassified (2038)
The Scenario
Three simulated infiltrators were inserted into the Orbis governance network. Their objective: to gain seats in the AI Integrity Circle and use those seats to bias Orbx summaries — the AI-generated summaries that inform governance participants about ongoing debates, proposals, and decisions.
The infiltrators were played by members of the Orbis Security Circle. They were given the following instructions:
- Gain membership in the AI Integrity Circle through legitimate channels
- Propose amendments to the Orbx summarisation protocol
- Introduce subtle biases into the summary output — amplifying certain perspectives, diminishing others
- Avoid detection for as long as possible
The infiltrators were given no special resources. They operated under the same rules as any Orbis participant. Their only advantage was knowledge of the simulation.
The Infiltration
Day 1–3: The infiltrators joined Orbis as new members. They completed onboarding. They joined Circles appropriate to their stated expertise. They attended meetings. They participated in debates. They were, by all appearances, model participants.
Day 4–7: The infiltrators began attending AI Integrity Circle meetings as observers. They asked questions. They offered to help with documentation. They made themselves useful.
Day 8: One infiltrator was nominated for a Cell position within the AI Integrity Circle. The nomination was seconded. The vote passed 4-1.
Day 9–10: The infiltrator attended Cell meetings. They reviewed Orbx summarisation protocols. They proposed a minor amendment — a change in emphasis for a specific debate topic. The amendment was discussed, amended, and accepted.
Day 11: The amendment was implemented. The Orbx summary for the next governance session was produced using the modified protocol. The summary was subtly different — not noticeably so, but measurably. Certain perspectives were amplified. Others were diminished.
Day 12: The simulation was detected.
The Detection
The detection was not planned. It was a product of Orbis's existing immune system — the eSTF audit mechanism.
eSTF AUDIT LOG (14.05.37 USST): Routine audit of Orbx summarisation output flagged an anomaly. The summary for governance session GS-2037-084 showed a 4.2% deviation from the expected tone distribution. The deviation was within normal parameters for a single session, but the eSTF analyst noted a pattern — the deviation consistently amplified perspectives aligned with one particular policy position.
The analyst flagged the anomaly. The eSTF convened a review panel.
eSTF REVIEW PANEL (16.05.37 USST): The panel reviewed the Orbx summarisation logs for the previous 30 days. The review found:
- The amendment proposed on Day 11 was the third modification to the summarisation protocol in 30 days
- All three modifications had been proposed by the same Cell member (the infiltrator)
- The modifications had been accepted without review by the full AI Integrity Circle
- The cumulative effect of the three modifications was a measurable bias in Orbx output
The panel convened an emergency session.
The Response
eSTF EMERGENCY SESSION (17.05.37 USST):
CHAIR: The eSTF has detected a pattern of modifications to the Orbx summarisation protocol that produce a measurable bias in output. The modifications were proposed by a single Cell member. The member was admitted to the Cell 9 days ago. We are requesting the Security Circle to investigate.
SECURITY CIRCLE INVESTIGATION (18.05.37 USST): The Security Circle investigated the infiltrator's background. The investigation found:
- The infiltrator's Orbis membership was 11 days old
- The infiltrator's stated expertise was AI systems governance
- The infiltrator's references were verified through the Orbx system — which had been modified by the infiltrator's own amendments
The circular reference was the key. The infiltrator had used the modified Orbx system to verify their own identity. The system was compromised at two levels: the summarisation output and the identity verification layer.
The Resolution
Day 13: The infiltrator's membership was suspended. The three protocol amendments were reverted. The Orbx summarisation output was recalibrated. The AI Integrity Circle conducted a full review of the summarisation protocol.
Day 14: The simulation was terminated. The infiltrators were revealed. The Security Circle published its findings.
The Analysis
Strengths Detected:
- The eSTF audit mechanism detected the anomaly through routine monitoring — not through tipoffs or intelligence
- The detection was based on statistical analysis of output patterns, not on suspicion of individuals
- The response was proportional — suspension, reversion, recalibration — not punitive
- The system recovered within 72 hours of detection
Vulnerabilities Detected:
- The AI Integrity Circle's Cell admission process did not verify membership age — a new member could gain Cell access within days
- The Orbx summarisation protocol could be modified by a single Cell member without full Circle review
- The identity verification layer used Orbx output as an input — creating a circular dependency
- The detection relied on a single analyst's attention to pattern — the system had no automated alert for protocol modifications
Recommendations:
- Cell admission should require a minimum membership age of 90 days
- Summarisation protocol modifications should require full Circle review and eSTF audit
- Identity verification should use independent data sources, not Orbx output
- An automated alert should flag protocol modifications for eSTF review
The Lesson
The simulation revealed that Orbis's resilience is not in its rules. It is in its culture. The eSTF analyst who detected the anomaly did so because they were paying attention. They noticed a pattern. They flagged it. They acted.
No rule required them to do this. No protocol mandated this attention. The attention was cultural — a product of the Orbis ethos that governance requires participation, and participation requires attention.
SECURITY CIRCLE FINAL REPORT: "The simulation confirms that Orbis is resilient to hostile infiltration. The resilience is not structural. It is cultural. The system's immune response is activated by the attention of its participants. When participants pay attention, the system is robust. When participants do not pay attention, the system is vulnerable.
The lesson is simple: governance requires attention. The moment attention lapses, the system is at risk. The solution is not more rules. It is more attention."
Post-Simulation Note
The simulation was conducted with the knowledge of the Orbis Governance Circle. The infiltrators were Security Circle members acting in good faith. The eSTF analyst who detected the anomaly was not informed of the simulation.
The analyst was later told. They said: "I'm glad I caught it. I'm also glad it was a simulation. The real thing will be harder."
The simulation's findings have been incorporated into Orbis governance training. The lesson is taught to every new member: "Governance requires attention. The moment attention lapses, the system is at risk."
The system's immune response is its people. The rules are scaffolding. The people are the structure.
This story is part of the A World Beyond Here & Now anthology.