About
The records
California lawmakers passed the landmark “Right to Know Act” in 2018, stripping away decades of secrecy and giving the public the right to see records about shootings by law enforcement officers and other serious uses of force, as well as law enforcement misconduct. Lawmakers strengthened the law in 2021, expanding the categories of misconduct cases disclosable to the public. Both these laws are found in California Penal Code section 832.7(b).
Since 2019, journalists, loosely organized as the California Reporting Project, have requested these records from nearly 700 law enforcement agencies, including police departments, sheriff’s departments, public school and university police forces, the California Highway Patrol, prisons, probation departments, district attorneys, the California Attorney General’s office, coroners, medical examiners, oversight boards and commissions, and other government agencies in California. Agencies have disclosed written investigative reports, findings, autopsy reports, written transcripts and videos of interviews of law enforcement officers, investigative hearings, photographs, and body-worn camera video in response to those requests.
Methodology
Through a state-funded project, the Berkeley Institute for Data Science (BIDS), UC Berkeley Journalism’s Investigative Reporting Program (IRP) and Stanford University’s Big Local News collaborated to build this database, called the Police Records Access Project. Journalists collected thousands of public records and, together with data scientists, organized those records. The team used emerging technologies, including generative AI, to build the database.
Cases are grouped into three broad categories of incidents disclosable under the law:
- Force: Police use-of-force incidents causing “great bodily injury” or death
- Shooting: Police shootings, fatal or not, including accidental and missed shots
- Misconduct: Sustained findings by agencies that an officer committed specific types of misconduct, including sexual assault, excessive force, dishonesty, prejudical action, unlawful arrest, and unlawful search
Some cases can belong to more than one category; all Shooting cases will also be listed in Force cases, for example.
The project uses large language models to help organize the trove of documents into cases, extract incident date and classify cases by type using information from all the textual files in each case. These uses of GenAI do not directly impact the information that is shown to users. For any search, only documents with a literal match to the search term(s) in the source text are returned.
To evaluate how well the AI models were able to find incident dates and classify cases, the team manually reviewed a subset of cases, sampled from every year in our collection across every type of agency that discloses records. Staff read files from those cases and identified dates and case types. This hand-crafted data set was used to compare against the model output. In addition, the team manually reviewed every case that machine learning categorized as misconduct. While the extraction review found results to be highly accurate, users should refer to the source files provided with each case.
Important Notes
Information may be incomplete
The database will be updated as we receive more records, analyze them, and extract more data.
This dataset does not include all use-of-force cases. The law requires government agencies to disclose only two types of use-of-force incidents:
- Discharge of a weapon at a person by police (excluding tasers, paintball guns, or other weapons using less lethal ammunition).
- Police uses of force resulting in death or “great bodily injury.”
The database also does not include all allegations of police misconduct. The law requires government agencies to disclose cases of “sustained” findings of sexual assault, excessive force, dishonesty, prejudical action, unlawful arrest, and unlawful search or where an officer was accused of those actions and left the agency before the investigation was completed.
Agencies are not required to release use-of-force records during active investigations.
Some records may be added after an investigation into a police shooting is closed and the agencies release the requested records.
Some information may appear in multiple files
Sometimes more than one agency may send records about the same case. A sheriff’s department, medical examiner and district attorney could have records about the same death. A local police department and the California Attorney General’s Office could review the same police shooting. California Highway Patrol officers and a sheriff’s deputy could respond to the same incident. When more than one agency sends records about a case, a search will produce separate documents from each providing agency.
Review the source documents
Search results should be carefully examined and compared with the PDFs of actual government records in the database, especially since some officers, witnesses, or others may share the same name. Officer names are ranked higher than others with the same name – bystanders or subjects, for example – in search results.
Searches also may not identify all relevant documents or cases. For example, a search for “taser” may not locate any cases when the government agency uses a different term for a taser, such as “electronic control weapon.” Users should not rely on search results alone and should review information in the government records in the database before arriving at any conclusions.
Redactions
These documents contain text or images some people might consider sensitive or disturbing. Where government agencies failed to redact personally identifying sensitive information from the disclosed government records, those records are not listed in search results.
In some instances, a case will display no files at all. This can happen when all the files in a given case contain sensitive information. We are working on a way to display more metadata about those redacted files. A case can also consist solely of multimedia documents - audio, images or video - which are currently not being processed. Those cases will not list files, however we plan to process multimedia in the future.
Who we are
The Police Records Access Project team is composed of members from the Community Law Enforcement Accountability Network, a coalition formed to obtain and distribute law enforcement records to the public. It is a nationwide collaborative effort involving journalists, data scientists, public defenders, community advocates, human rights defenders, and First Amendment lawyers, including:
ACLU of Southern California
ACLU of Northern California
Bay Area News Group / Southern California News Group
Berkeley Institute for Data Science
Big Local News at Stanford University
CapRadio
EPIC Data Lab
Human Rights Data Analysis Group
Innocence Project
The Investigative Reporting Program at the UC Berkeley School of Journalism
LAist
KQED
Los Angeles Times
National Association of Criminal Defense Lawyers
UC Berkeley School of Law
UC Irvine School of Law
Feedback
We’re constantly striving to improve our processes and methods. Please contact us if you have any suggestions.
If you have questions or feedback, click here.