Submit a case

AI Incident Law is a curated public-record corpus. Contributions are welcome from anyone -- law school clinics, in-house counsel, journalists, researchers, opposing counsel who watched a case turn on AI-related conduct. The dataset does not favor any particular product or firm, and admission is governed by stated criteria, not editorial taste.

Open a record-submission issue Open a pull request

What we accept

A candidate record is admitted to the included bucket when all of the following hold. These are the same criteria the steward applies to her own submissions; they live in INTENT.md as the authoritative statement.

1
Public legal or regulatory matter directly involving AI-related conduct, output, or use. Filed proceedings, regulatory actions, settlements, judgments, consent decrees, formal investigation disclosures. Press coverage alone is not sufficient.
2
At least one primary or reliable secondary public source. Court filings, agency orders, tribunal opinions, official press releases, or stable archival mirrors (CanLII, CourtListener, RECAP, official gazettes).
3
Required fields present and consistent. Jurisdiction, parties, AI-relevance description, source URLs, dates. See the data schema for exact field shapes.
4
AI-relevance is concrete, not speculative. The matter must turn on AI-related conduct in a way a court or regulator named, not merely that "AI was in the room." Generic AI ethics statements without legal or regulatory action attached are out of scope.

Where your submission goes

included

Meets all four admission criteria. Renders on the main site, appears in the Obligation-First API, and is fetchable by the public MCP.

review

In scope but needs verification, primary-source strengthening, or schema cleanup. Visible in the dataset but flagged. Contributors can iterate from here.

global

Non-US matters that need translation, jurisdiction-specific interpretation, or additional sourcing. Held here until the editorial workflow can do them justice.

The submission process

Option 1 -- open an issue (recommended for first-time submitters)

Open a GitHub issue with the data label. Include:

matter name and jurisdiction
primary source URL (court filing, agency order, tribunal opinion)
one-sentence statement of why this is AI-related, with the specific conduct in question
which admission criterion you believe the matter satisfies

The steward will respond with a triage: admission, request for more sourcing, queue into review, or out of scope with a brief explanation. No silent rejections.

Option 2 -- open a pull request (for contributors comfortable with JSON)

Edit data/data.json directly. Add your record to the appropriate bucket (included, review, or global) following the field shapes documented in docs/data-schema.md. Run npm run build and npm run check locally before opening the PR so URL policy and validation pass.

Read CONTRIBUTING.md for the exact editing rules.

Anchoring to obligations

If your submitted matter has resolved (has a disposition -- settled, ordered, sanctioned, etc.) and you can identify which EveryAILaw obligation(s) the breach implicates, include them as obligation_first_anchors. This is what makes the matter traversable from the PAICE legal graph. If the matter is still pending, leave the field off; anchors are only valid on resolved records.

What we do not accept

Speculative AI-risk commentary or ethics statements without legal or regulatory action attached.
Internal corporate disputes that have not surfaced publicly.
Reputational controversy that has not produced a filing or formal action.
Private filings, leaked material, or anonymous tips. Every record must resolve to a public source.
Records under non-disclosure or otherwise barred from public reference.

Attribution and licensing

Submitted records are published under CC BY 4.0. The dataset preserves contributor attribution in git history. By submitting you confirm that the underlying sources are public and that you have the right to reference them.

Reach the steward

For questions, scoping conversations, or sensitive submissions that need a private channel before going public, contact PAICE.work PBC.