Human Rights Access Lab · Clear Mirror / 明鏡計劃
Active · public research and prototype development
When the system says find a lawyer—and there is no lawyer.
We document where human-rights access fails, audit access to the underlying decisions, and build a Case Pathfinder that keeps every result connected to the source, the search and the next procedural step.
- Program status
- Active · public research and prototype development
- Record updated
- October 13, 2026
Why this program exists
A right can exist on paper while the route to use it has collapsed.
- 01
A person encounters conduct that may engage human-rights protection.
- 02
They cannot find a lawyer willing to take the matter, or cannot afford repeated private consultations.
- 03
The system still expects them to manage deadlines, procedure, evidence and legal authorities.
- 04
They are left to research alone.
- 05
General-purpose AI can help, but it can invent authorities, omit relevant decisions and attach a citation to the wrong proposition.
- 06
The public alternative must connect conclusions to original decisions, show what is missing and preserve a reproducible research trail.
The lived starting point
$364 for one hour. About forty-five minutes spent telling the story again.
Contemporaneous emails confirm the price, the one-hour booking, the advance-document request and the completed payment. The participant recalls that about forty-five minutes were spent retelling a chronology already sent in writing, without sustained help at the end. Five months later, after sending the chronology again, the same firm declined to assist. Other routes had no employment appointment, offered only 15–30 minutes, could not review a 59-page record, or said the issues were too complicated and nuanced and pointed back to paid counsel. The public account keeps this causal chain while removing names, medical information, file numbers and details that could affect live proceedings.
Read the anonymized journeyOpen research desk
Follow the evidence, the method and the build.
The Lab is published as a set of linked working records. Each page says what is known, what still needs to be verified and what will be built next.
Build roadmap
Prove the research chain first. Add funded inference where it earns its place.
- 1
Now
Authority and access audit
Verify the BCHRT corpus, source metadata, reuse terms, correction history and lawful machine-access route. Publish gaps and adverse findings.
- 2
First public build
No-cloud-token foundation
Guided facts, dates, ground, forum and stage; authorized decisions indexed with SQLite FTS5/BM25; deterministic procedural rules; citations, excerpts and exports.
- 3
Funded layer
Multilingual, secure inference
Natural-language intake, semantic retrieval, decision comparison and source-linked plain-language summaries, with privacy controls and independent evaluation.
- 4
Ongoing
Maintenance and accountability
Corpus updates, regression tests, correction notices, community testing and public reporting on accuracy, omissions and unresolved access barriers.
Current evidence changes the claim
Search+ is useful. It is not a public corpus or a durable access right.
Search+ is open to free myCanLII accounts, but currently allows ten generated queries and four analyses per day; each analysis reads only the first 25 results. Even an unlimited version would leave the public dependent on one intermediary’s interface, terms, selected model and continued availability. The unresolved question is who must provide a stable, authorized route to retrieve, preserve, reuse and computationally analyze public tribunal decisions.
See the decision-access auditWork with the Lab
We are looking for technical, legal-aid, research and funding partners.
The useful conversation starts with the corpus, the user journey, the evaluation plan and the limits—not a vague request for AI tokens.