Social Media

Light
Dark

JusticeText taps AI to transcribe evidence for public defenders

While pursuing their computer science studies at the University of Chicago, Devshi Mehrotra and Leslie Jones-Dove were motivated to develop technology that catered to the requirements of communities historically marginalized by law enforcement. Their inspiration came from local public defenders who conveyed their overwhelming workload, which included processing hours of jail calls, body camera footage, and other recorded evidence.

To put it in perspective, an average officer’s body camera records approximately 32 files, 7 hours, and 20GB of video per month at a 720p resolution. When you multiply these numbers by the hundreds or even thousands of officers in a police force, the volume becomes staggering.

Mehrotra explained in an email interview, “While body cameras and similar devices are essential for holding law enforcement accountable and ensuring the best possible defense, they exacerbate the workload challenges faced by public defenders, who are dealing with caseloads 3 to 10 times the recommended amount.”

In response, Mehrotra and Jones-Dove founded JusticeText, one of the startups participating in the Startup Battlefield 200 at TechCrunch Disrupt 2023. JusticeText is designed to automatically transcribe body camera footage, interrogation videos, and other types of evidence for public defenders, allowing attorneys to create time-stamped notes, generate video clips, and share evidence with their colleagues.

JusticeText also offers a feature that provides concise summaries of each piece of uploaded evidence. Moreover, its ChatGPT-powered tool, MirandaAI, enables attorneys to ask open-ended questions about their discovery, such as “Were there any implicit promises made to the client?” or “What is the timeline of events described by the witness?”

Mehrotra emphasized, “Public defenders owe their clients the best possible defense. However, digital discovery review consumes an increasingly large share of limited resources. Saving time on discovery review frees up critical resources that can be devoted to working on the case and building client relationships. From a technical standpoint, it also means fewer IT resources are needed for technical problem-solving, like handling videos in unique proprietary formats or creating video clips for an upcoming trial.”

While these benefits sound promising in theory, some public defenders may hesitate to upload evidence to a platform like JusticeText due to concerns about data security and potential misuse of uploaded data, such as training JusticeText’s AI models.

Additionally, there are concerns about the accuracy of JusticeText’s transcriptions, especially when dealing with different accents and languages. The summaries provided by JusticeText may lack nuance, as AI is not known for its ability to grasp subtleties.

Mehrotra sought to address these concerns by assuring that JusticeText generates summaries only for transcripts that meet a certain confidence threshold. Users also have the option to redact confidential information from transcripts. Regarding privacy and data storage, JusticeText stores data on secure cloud servers and encrypts it during both transit and at rest. Moreover, JusticeText refrains from using uploaded data for model training unless explicit permission is granted.

These measures have been sufficient to win over clients. Since closing its $2.5 million funding round with participation from Bloomberg Beta, True Ventures, LinkedIn co-founder Reid Hoffman, and former Stockton, California, mayor Michael Tubbs, JusticeText has expanded its customer base to include statewide public defender systems in Massachusetts and Kentucky, in addition to over 100 public defender agencies, nonprofit service providers, and private practice criminal defense firms across the United States. The company now boasts $1 million in annual recurring revenue, a significant increase from September 2022 when they had 50 to 60 partnerships.

Looking ahead, JusticeText plans to introduce Spanish-to-English translation capabilities and support for “multi-language” recordings, where speech is in both English and Spanish. Mehrotra emphasized the importance of careful implementation given AI’s imperfections as a translator. The company also plans to expand its team from seven employees to around 10 in the coming year, with a focus on enhancing its marketing and communications efforts.

Mehrotra concluded, “The pandemic has created a backlog in the criminal court system, affecting our end users. Despite the challenges in reaching key decision-makers, these circumstances create a compelling reason for change.”

Leave a Reply

Your email address will not be published. Required fields are marked *