Civic AI answers to the people it affects — alignment by public process: community-authored safeguards, public accountability, and bounded local systems.
This repository contains the public site for the 6-Pack of Care, a governance framework by Audrey Tang and Caroline Green at Oxford's Institute for Ethics in AI.
The core idea is simple: instead of asking a small group of developers to define "aligned" for everyone, the 6-Pack asks who is affected, who can contest decisions, what gets logged publicly, and how systems stay local and reversible. The 6-Pack treats AI not as a sovereign optimiser but as a bounded local steward, or Kami.
| # | Pack | Tronto phase | Core question |
|---|---|---|---|
| 1 | Attentiveness | Caring about | What do the people closest to the pain notice that we're missing? |
| 2 | Responsibility | Taking care of | Who is accountable, with what authority, and what happens if they fail? |
| 3 | Competence | Care-giving | Does the system demonstrably work — audited, explainable, safe-to-fail? |
| 4 | Responsiveness | Care-receiving | Can those affected correct the system, and does correction actually change it? |
| 5 | Solidarity | Caring with | Does the ecosystem structurally reward cooperation over lock-in? |
| 6 | Symbiosis | Kami of Care | Is the system bounded, sunset-ready, and incapable of imperial creep? |
Packs 1 – 4 form Tronto's feedback loop. Pack 5 (from Caring Democracy) ensures the loop operates within democratic commitments to justice, equality, and freedom. Pack 6 is Tang and Green's addition: the meta-level guardrail that keeps care local, bounded, and provisional.
Just heard about Civic AI? The friendliest way in is the website — civic.ai — which renders all of this with illustrations and audio. Three steps:
- Get the idea. Read the Manifesto — the whole argument in Audrey Tang's own words.
- Meet the six principles. Skim the six packs above: plain-language tests for AI a community can actually trust.
- See it work. AI Alignment Cannot Be Top-Down tells how Taiwan answered a wave of AI-powered scam ads.
Already know what you're after?
- Policy & governance — Manifesto, FAQ, and AI Alignment Cannot Be Top-Down.
- Builders & engineers — Pack 3: Competence, Inside the Kami, and Measures.
- Civic & community practice — Pack 1: Attentiveness, Pack 4: Responsiveness, and the podcast.
civic.ai — bilingual (British English/Traditional Mandarin) static site.
Built with Eleventy v3 and Bun:
bun install # install dependencies
bun run dev # local dev server at http://127.0.0.1:8080
bun run build # production build → ./docs/| Path | What it is |
|---|---|
*.md (top level) |
Page content, British English. One file per page. |
tw-*.md |
Traditional Mandarin counterpart of each English page (keep in parity). |
_layouts/ |
Page templates: default.html (home), chapter.html (book pages, prev/next nav), conference.html (conference page). |
_includes/ |
Shared template partials pulled into the layouts (page-head, site-logos, site-nav, scripts, footer, seo-head). |
_data/ |
Global data (site config, comics, OpenClaw bootstrap, Polis report). |
img/, fonts/, audio/ |
Static assets, passthrough-copied to the build. |
styles.css |
All site styles (mobile-first; uses CSS custom properties). |
eleventy.config.js |
Build config: passthrough rules, Markdown tweaks, filters. |
specs/ |
Internal design & implementation docs (not published; see .eleventyignore). |
docs/ |
Generated build output — never edit by hand. |
Pull requests are welcome. By contributing, you agree to release your work under the CC0 1.0 Universal public domain dedication.
When editing content, maintain parity between English (*.md) and Traditional Mandarin (tw-*.md) variants. English files use a spaced single em dash (—); tw-* files use a double em dash (——, no spaces). Reuse the project's locked Traditional Mandarin terminology rather than coining new translations.
Part of the Accelerator Fellowship Programme, Oxford Institute for Ethics in AI.