[uk ai resilience] UK AI Governance: recent-change risk review (2026-09-17 to 2026-09-24) #63216
Closed
Replies: 1 comment
|
This discussion has been marked as outdated by UK AI Operational Resilience. A newer discussion is available at Discussion #63464. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Executive summary
7-day recent-change sweep (2026-09-17 → 2026-09-24, 109 commits, 44 security-signal commits) found no new untracked risk: all 272 open code-scanning alerts and 0 open secret-scanning alerts already have open, dated
uk-ai-resiliencetracking issues from prior governance cycles. The governance loop's classification step is therefore healthy. The one persisting operational gap is in recovery/closure hygiene: three alerts (#669, #668, #667) remain open in code scanning despite tracking issues (#57472, #54037, #59491) asserting the underlyinggo/bad-redirect-checkfinding is a false positive — and source inspection confirms the proposed shared safe-path-validation helper has still not been implemented. No new issues are created this run to avoid duplicating the 62 open security issues already covering every current finding; this report instead re-confirms and dates the evidence for the existing dismissal-hygiene tracking issues (#57982, #59490) and the redirect-check remediation issue (#59491).Note: the
asset-tier-classifierandcontrol-verifiersub-agents failed twice on an infrastructure model-access error (gpt-5.4-mini not accessible, unrelated to this repo's content) and could not be re-run; findings below were compiled directly from the pre-computed evidence files and direct source/issue inspection, so confidence on the asset-graph and control-verification sections is slightly reduced (recorded as a limitation, not fabricated).Asset graph summary (recent-change scoped)
pkg/workflow/*(compiler: mcp_*.go, compiler_yaml_ai_execution.go, graders_config.go)pkg/cli/*(add_package_manifest*.go, project_command.go, logs_orchestrator_stdin.go).github/workflows/*.lock.yml(254 generated files)Dockerfilealpine:3.24base (#61166, alert #940)scripts/ensure-docs-slide-pdf.jsREADME.md.changeset/*(694 files).github/CODEOWNERS(absent)Ownership is inferred from commit authorship only (no CODEOWNERS): recent window dominated by
Copilot/copilot-swe-agent[bot]commits co-authored bypelikhan, plusgh-aw-bot/github-actions[bot]automation merges.Control verification gaps
.github/CODEOWNERS(confirmed via directls); tracked in [uk-ai-resilience] Missing .github/CODEOWNERS for security-sensitive compiler/CLI paths (Tier B) #61637 (Tier B)..changeset/*files and consistent co-authorship trailers indicate disciplined change tracking, but the volume makes manual review of each security-relevant diff impractical without owner routing.pkg/workflow, indicating baseline observability instrumentation exists for the compiler pipeline.isSafeRelativePath-style helper exists inpkg/cliorpkg/workflow. This is the same gap raised in [uk-ai-resilience] Alert-dismissal hygiene gap: 3 CodeQL alerts remain open despite closed/remediated tracking issues (Tier C) #57982 and [uk-ai-resilience] Alert-dismissal hygiene gap persists across CodeQL tracking issues (Tier C) #59490, now confirmed still present ~3 weeks later. One partial improvement: alert #653 (previously flagged as stale in [uk-ai-resilience] Alert-dismissal hygiene gap: 3 CodeQL alerts remain open despite closed/remediated tracking issues (Tier C) #57982) is no longer in the open-alerts list, suggesting some dismissal hygiene progress has occurred.Risk-scoring table
Remediation queue with SLAs (all already tracked — no new issues opened)
.github/CODEOWNERSfor compiler/CLI security pathsException register
No new exceptions proposed this run. Existing accepted-risk candidate (pending, not yet formalized): pinned-version-only installs for AI engine CLIs as a temporary mitigation for the npm/uv-pip determinism cluster, per #61636 — requires an explicit expiry date and mitigation plan before being treated as a governed exception; currently informal.
Operational metrics baseline
Warning
Firewall blocked 1 domain
The following domain was blocked by the firewall during workflow execution:
o205451.ingest.us.sentry.ioTo allow these domains, add them to the
network.allowedlist in your workflow frontmatter:See Network Configuration for more information.
All reactions