Content exclusions without proof
GitHub honors repository-level content exclusions. Your auditor asks how you prove an exclusion was in effect at the moment of a specific completion. Admin settings alone are not evidence.
Copilot runs in GitHub's cloud. You cannot proxy it the way you proxy other coding tools. What you can do is bring Copilot activity into your governance plane, pair it with a tamper-evident audit trail, and enforce one policy across every coding AI in your org.
Copilot Business and Copilot Enterprise ship with content exclusions and admin logs. They do not ship with tamper-evident evidence, unified cross-tool policy, or auditor-ready export. Raidu adds those layers.
Admins configure Copilot exclusions and get a log. Regulators ask for tamper-evident evidence, cross-tool consistency, and retention guarantees. That gap is what Raidu closes.
GitHub honors repository-level content exclusions. Your auditor asks how you prove an exclusion was in effect at the moment of a specific completion. Admin settings alone are not evidence.
Copilot's audit log is pulled from GitHub on demand. It is not chained so that changes show, and it is not retained by you. If the record is disputed, the trust root is GitHub, not you.
Most teams run Copilot plus Cursor plus an Anthropic or OpenAI client. Each has its own admin panel. Security writes a policy once and enforces it three times. Drift is inevitable.
Auditors ask for one evidence bundle covering AI coding assistance across the org. Copilot produces some. Other tools produce some. Nobody produces one record across all of them. That is the gap.
Raidu does not intercept Copilot's outbound traffic. GitHub does not allow that. Raidu complements Copilot with the three things GitHub does not provide: cross-tool policy, tamper-evident independent evidence, and auditor-ready export.
Define one policy for AI coding assistance. Raidu enforces it at runtime for Cursor, Cline, Continue, Claude Code. For Copilot, Raidu mirrors the rules into GitHub's admin settings via the Copilot admin API where supported, and flags drift.
Raidu ingests Copilot's usage logs via the GitHub API on a regular schedule, rebuilds the hash chain on your side with SHA-256 hashes, and persists it to WORM with 10-year retention. Your evidence is yours.
Every configured exclusion is hashed and recorded at the time it is set. When an auditor asks whether exclusion X was in effect on date Y, Raidu returns a tamper-evident timeline.
Export a single evidence bundle covering Copilot and every other AI coding tool in the org. Auditor-ready. Regulation-mapped. Tamper-evident, inside your own cloud account.
Point Raidu at your GitHub org. Raidu pulls Copilot usage, rehashes, signs, and layers it under your unified policy.
{
"connector": "github-copilot",
"org": "acme-corp",
"auth": "github-app",
"ingest": {
"usageLogs": "hourly",
"contentExclusions": "on-change"
},
"policy": "coding.eng.v7",
"hash": "sha256",
"retain": "10y"
}
// Every Copilot session becomes a tamper-evident record on your side, under your policy.Thirty minutes. We connect a sandbox GitHub org, ingest real Copilot logs, sign them under test keys, and export a demo evidence bundle.