

The future of AI and the law
9 snips Jul 11, 2025
Daniel Ho, a law professor at Stanford specializing in AI and legal documentation, discusses the transformative potential of artificial intelligence in the legal field. He highlights the risks associated with AI hallucinations in legal queries, stating over 80% can be inaccurate. Ho shares a groundbreaking project in California where AI was used to swiftly identify and remove racist property covenants from laws. He emphasizes AI's role in reducing bureaucratic inefficiencies, paving the way for legal reform and increased access to justice.
AI Snips
Chapters
Transcript
Episode notes
AI Legal Hallucinations Are Common
- Large language models hallucinate legal facts 60 to 80% of the time without retrieval augmentation.
- Even augmented systems still hallucinate 20 to 33% of the time, so caution is needed.
AI Accelerates Redacting Racist Covenants
- California mandated counties redact racist property covenants from deed records, a massive task.
- AI reduced the identification and redaction from years to days, accelerating the process dramatically.
Multimodal AI Improves OCR Accuracy
- Multimodal AI models outperform traditional OCR in accuracy and cost.
- This leapfrogging allows better digitization pipelines for legal documents like deed records.