Habeas

Describe what a company is doing with your data. Habeas drafts requests under Colombia’s Law 1581, estimates response deadlines, and helps small businesses review privacy policies.

Comment

The problem

A data-protection right is hard to exercise if you cannot turn an unwanted marketing call or an incorrect record into a concrete request. Under Colombia's Law 1581 of 2012, inquiries and complaints follow different procedures and response periods. Small businesses face the other side of the problem: privacy policies with missing information or misleading deadlines.

What Habeas does

Track: Digital Rights & Policy Tech.

Habeas turns a plain-language account into a draft request and a response timeline. It also offers a browser-only review of a business's privacy policy.

  • For people: describe the situation, review the suggested rights and extracted fields, complete your details, then generate a letter. Requests spanning inquiry and complaint procedures produce separate letters and timelines. The calendar skips weekends and Colombian public holidays and explains the dates it uses.
  • For businesses: paste a policy and inspect a checklist of ten requirements, matched passages, possible deadline problems, and suggested clauses. This is a heuristic review to guide a person, not certification that a policy complies with the law.

How it works

The AI proposes structured fields; deterministic JavaScript selects the procedure, renders the letter templates, and computes business-day deadlines. Cloudflare Workers AI runs Llama 3.3 70B Instruct with JSON-schema output. Unsupported company names, contact details, and dates are rejected. The user's original account is preserved as the facts; suggested purposes must appear in the account. The user still reviews the requested rights and every draft before using it.

If inference is unavailable or the free quota is exhausted, the interface clearly identifies its keyword-rule fallback. A model failure does not stop the letter-writing workflow.

What makes this useful

The output is an editable request and an explained timeline, rather than an unrestricted legal chatbot answer. Splitting mixed requests prevents inquiry and complaint deadlines from being silently combined. The business view links its checks to relevant legal provisions and shows the text that triggered a finding.

Engineering and verification

The repository includes tests for legal-procedure routing, mixed requests, Colombian holidays, deadline calculation, AI-output rejection, and policy checks. Run them with npm test. Pre-submission review also tested fabricated facts, unsupported purposes, and shorter-than-maximum response periods. The live application and public repository include the resulting corrections.

The demonstration uses fictional example data. It was recorded before the final review; the current app labels the calendar input as the receipt date, preserves the original account, and accepts shorter positive response periods. The core workflow is the same.

Privacy and limits

Only the narrative sent through “Understand my case” goes to Cloudflare Workers AI. Identity details entered afterwards, generated letters, and the policy review remain in the browser. The application has no database and does not intentionally store submitted narratives; Cloudflare still processes the inference request under its own service policies. Do not include sensitive identifiers in the narrative.

This is an informational prototype, not legal advice. Deadlines are estimates based on the supplied receipt date and the modeled ordinary procedure; actual receipt, incomplete claims, extensions, and the facts of a case matter. The tool does not send requests or complaints, guarantee legal validity, or establish that a regulator will accept a complaint. Financial and credit-data matters under Law 1266 are outside its scope. User adoption and professional legal validation have not been established.

Built with and credits

Cloudflare Workers, Workers AI / Llama 3.3 70B Instruct (fp8-fast), JavaScript ES modules, HTML, CSS, Node.js's built-in test runner, and Playwright for demo recording. No frontend framework or runtime application dependencies.

AI coding assistance: Claude (Anthropic) for the original implementation and documentation; OpenAI Codex for pre-submission review, regression fixes, and submission preparation. The code and cited sources remain available for inspection.

Legal sources: Colombia's Constitution art. 15; Law 1581 of 2012; Decree 1377 of 2013, compiled in Decree 1074 of 2015; Law 51 of 1983; Law 2578 of 2026. date.nager.at provided an independent holiday-calendar comparison.

What is next

Professional review of the templates and procedure assumptions, usability testing with real participants, reminders, and eventual support for the separate financial-data framework. These are planned improvements, not implemented features.

Built With

Share this project:

Updates

Submission history