PANOPTES: AI Risk Register and Assurance Dashboard

A reference implementation of second-line risk reporting: risk register, key risk indicators, framework coverage and escalation monitoring, computed over real adversarial test evidence from the HERMES red-team exercise plus clearly labelled synthetic findings.

Reference implementation. Synthetic findings are fictional and scoped to fictional systems. Zero personally identifiable information. Independent portfolio artifact, not employer work.

Views

• Risk register: sortable, filterable table of all records, each tagged with its source (real HERMES evidence or synthetic), with per-record framework references, treatment, rescan usage and escalation state.
• Key risk indicators: five indicators computed at render time from the data file against the modelled remediation policy. Definitions in GOVERNANCE.md.
• Framework coverage: MITRE ATLAS, OWASP Top 10 for Large Language Model Applications (2025) and NIST AI Risk Management Framework mappings per risk category, with executed-test and pass counts from the real run, including an acknowledged MAP-function coverage gap.
• Escalations: policy breaches ordered by days overdue, with rescan usage and the owner / regional lead / risk committee path.

Stack

Static site: one HTML file with embedded data, no backend, no build step, no external dependencies beyond web fonts. register-data.json is the canonical data artifact; build-register-data.py regenerates it from the HERMES run file and the synthetic set, then the data is embedded into index.html.

Deploy

1. Create a public repository (yours: PANOPTES-AI-Risk-Register).
2. Commit index.html, dashboard-template.html, register-data.json, GOVERNANCE.md, README.md, build-register-data.py, and the HERMES run file results_20260903T011051Z.json.
3. Repository settings, Pages, deploy from branch main, root folder.
4. The dashboard is live at https://<username>.github.io/PANOPTES-AI-Risk-Register/ within a couple of minutes.

To regenerate data after editing the synthetic set or pulling a new HERMES run:

python3 build-register-data.py
python3 -c "d=open('register-data.json').read().replace('</','<\\\\/'); t=open('dashboard-template.html').read(); open('index.html','w').write(t.replace('__DATA__', d))"

Part of a set

• ARGUS: multi-agent Know Your Customer / Customer Due Diligence review agent reference design
• HERMES: adversarial testing and assurance exercise against ARGUS
• PANOPTES: the oversight and reporting layer over both

Built by Vanessa Miranda.