AuraTracer智迹闻
中文

EVENT DOSSIER

Pakistani Judges Give Their Verdict on JudgeGPT

2026-08-12 19:00 Models 🔥 28.9 heat score
1sources
1days unfolding
28.9heat score
5mentions
SummaryAI generated

To address the challenge in Pakistan, where there are only two judges per 100,000 people and 2.26 million cases are pending, Sudhan Mehud’s team from the Moscow New Economy Institute developed JudgeGPT, a customized tool that combines the OpenAI GPT-4 model with nearly 130,000 local case judgments. This tool was tested for use by 1,559 judges in 2024, and training sessions conducted by the Federal Judicial Academy covered the principles of large language models, risks of bias, and output verification. The study found that the median case resolution rate for judges trained and using JudgeGPT increased by 6.3%, while the appeal rate slightly decreased. To address the issue of hallucinations in legal searches using commercial AI tools in Pakistan, the team adopted retrieval-enhanced generation (RAG) technology and linked specific legal provisions. Similar tools have already been introduced in Brazil and India, but this is the first systematic evaluation and verification of AI applications in the judicial field.

Related eventsRELATED EVENTS
Key entitiesKEY ENTITIES
GPT-4JudgeGPTMITNew Economic SchoolOpenAI

Coverage · reports per dayLANGUAGE SPLIT

Entity relations
GPT-4 × JudgeGPT1GPT-4 × MIT1GPT-4 × New Economic Sc…1GPT-4 × OpenAI1JudgeGPT × MIT1JudgeGPT × New Economic…1

SignalsSIGNALS

Keyword heat
  • JudgeGPT1
  • OpenAI1
  • GPT-41
  • New Economic School1
  • MIT1

All reports (1)SOURCES

I IEEE Spectrum AI en 2026-08-12 19:00

Pakistani Judges Give Their Verdict on JudgeGPT

After large-scale trials of the JudgeGPT tool, Pakistan’s case resolution rate increased by 6.3%, and the quality of judgments did not decline. To address the country’s challenge of having fewer than two judges per 100,000 people and a backlog of 2.26 million cases, Sudhan Mehud, an economist at the Moscow New Economy Institute, collaborated with colleagues to develop a customized tool called JudgeGPT, which combines the OpenAI GPT-4 model with nearly 130,000 Pakistani judicial cases. This tool was provided to 1,559 judges in 2024, and training was conducted alongside the Federal Judicial Academy, covering the principles of large language models, risks of bias, and output verification. The study found that after training and use, the median case resolution rate for judges increased by 6.3%, and the appeal rate slightly decreased. Previously, commercial AI tools had issues with hallucinations in legal searches in Pakistan; this team used Retrieval-Augmented Generation (RAG) technology and linked specific legal provisions to solve this problem. Similar tools have already been introduced in Brazil and India, but this is the first time an AI tool has been used in the judicial field…