I am a policymaker/researcher. How can I use STARA?
STARA is currently available in closed beta, and we’re excited to put the system in the hands of policymakers and researchers. If you’re interested in using STARA, please feel free to reach out to us at reglab [at] law [dot] stanford [dot] edu. We are not able to respond to all inbounds, but it is particularly helpful if you can let us know who you are, what jurisdiction you’re interested in, and what kind of reform/research effort you’d like to undertake.
What jurisdictions does STARA cover?
STARA’s architecture is designed to be jurisdiction-agnostic. The system currently supports the United States Code, the Code of Federal Regulations, San Francisco’s Municipal Code, and the statutory codes of all 50 states. Other legal codes can be added to STARA where there exists well-structured markup such as XML or HTML.
What makes STARA more effective than general purpose systems like ChatGPT?
STARA builds on the same underlying large language models used by services like ChatGPT! However, by default, these systems perform poorly on complex statutory research tasks for a few reasons:
- Unusual and diffuse statutory language: Statutory provisions have little in common with traditional prose. They contain highly precise language, terminology that may be defined in distant sections, and a web of non-obvious interdependencies. In many cases, individual subsections are nothing more than sentence fragments, understandable only with reference to language contained in a parent provision. Via its preprocessing pipeline, STARA supplies LLMs with the nonlocal context necessary—definitions, cross-references, headings, etc.—to make sense of provisions.
- Lack of systematicity: Most AI research systems are meant to gather some number of relevant documents and then summarize them. But STARA is meant for systematic research, where one must locate all statutory language meeting some criteria. It therefore takes a fundamentally different approach from other research systems—batch annotation followed by aggregation, rather than a single thread of research.
Why are statutory surveys important?
Statutory surveys are essential for legal reform. By systematically identifying all provisions relevant to a legal issue, they reveal duplicative, conflicting, or outdated laws that desperately need revision. Our recent partnership with San Francisco's City Attorney provides a compelling example: using STARA to identify reporting requirements led to legislation proposing to eliminate or amend 36% of all reporting obligations, freeing up staff time during an era of budget constraints.
At the federal level, the bipartisan “Count the Crimes to Cut Act of 2025” exemplifies long-standing interest in the systematic enumeration of federal criminal offenses. When the U.S. Department of Justice last attempted to count federal crimes in 1982, its team of lawyers gave up after two years of work with only a rough guess. The official who oversaw the effort later remarked that one could “die and [be] resurrected three times” and still not know the true number. STARA has changed that: we release a dataset of 1,983 federal criminal provisions, which is the most comprehensive enumeration of federal criminal offenses to date. And unlike previous efforts, every provision in our dataset is annotated with rich metadata, including offense descriptions, penalties, and detailed explanations of the legal reasoning behind the classification.
Beyond reform, statutory surveys serve critical functions across government and legal practice—helping agencies and private entities understand their obligations, enabling citizens to hold officials accountable, and providing researchers with empirical evidence on fundamental questions in political science and administrative law. However, traditional statutory surveys are extraordinarily labor-intensive, requiring researchers to manually read and annotate thousands of provisions. Tools like STARA transform this process by automating the tedious work while maintaining accuracy, making comprehensive legal reform and rich empirical research feasible.