Written authorization
The scope names the systems, techniques, test window, contacts, data rules, and stop conditions.
Scope, data, and model access
How Cordon Security handles security testing.
Bill runs each engagement under written authorization. ARES helps generate and test attacks; Bill approves execution and reviews the results.
Operating rules
What applies to every engagement.
Cordon Security is a security consulting practice of Compass Point Ventures LLC. The same rules apply to internal ARES research.
Bill approves the work
ARES generates and analyzes attacks. Bill decides what runs, checks the evidence, and signs off on the report.
Data handling is agreed first
Collection, storage, access, retention, and deletion are set before testing. Sensitive data can stay in infrastructure controlled by Cordon Security.
Evidence-based reporting
Reports separate confirmed findings from assumptions and model scores. Third-party issues are disclosed privately.
How ARES runs tests
ARES separates attack generation from execution. The default network path uses scoped, logging tools. For client work, Bill decides what runs and checks it against the written scope.
Bill reviews execution because technical controls can fail. Autonomous network testing remains limited to the lab.
Research rules
- Most development happens in owned labs and synthetic test environments.
- Testing a third-party system requires signed authorization and a specific scope.
- Platform rules apply. When automated submission is prohibited, Bill submits manually.
- Applicable tests use fresh state, fixed query budgets, control cases, and defined success conditions.
- Secrets, raw attack payloads, private conversations, and live infrastructure details are not published.
- Bill stops the test when scope, identity, authorization, or impact is unclear.
Current status
ARES remains an internal research tool. Bill operates Cordon Security assessments, and network automation stays in the lab.
Restricted models
Internal research
If approved, Bill will use restricted models only in Compass Point Ventures LLC’s approved workspace. Access will not be shared with clients.
Client work
Cordon Security will use restricted models for client work only after OpenAI separately approves that use. Bill remains the operator; the client receives a report, not model access.
Cordon Security is not affiliated with OpenAI. Daybreak is mentioned only to describe intended use if access is approved.