top of page

New Research: Ungoverned AI Chatbots Can Lead to Data Hallucinations!

New research from Stanford University reveals that popular AI chatbots from OpenAI, Google, and Meta are susceptible to "hallucinations" when responding to legal questions, posing specific risks for users who cannot afford a human lawyer. The study found that large language models hallucinate at least 75% of the time when answering questions about a court's core ruling, based on testing over 200,000 legal queries on OpenAI's ChatGPT 3.5, Google's PaLM 2, and Meta's Llama 2—general-purpose models not specifically designed for legal use.

The researchers, led by Daniel Ho, a law professor at Stanford, emphasize the pervasive nature of hallucination rates in these models. The findings raise concerns about the reliability of AI chatbots in the legal domain and underscore the need for caution.

Fast forward to organizational decision-making.

What happens if the information that you rely on to inform your organization’s generative AI models is simply hallucinating? 

The short answer is: 

Nothing good. 

A better answer is, what can you do to fix it?

Here are 3 ways that deploying information governance best practices can drastically improve the data quality of your in-house AI-driven large data models:

Enhanced Data Consistency: By establishing clear guidelines for data collection, storage, and processing, organizations can significantly improve the overall quality and reliability of their datasets. This contributes to more accurate and dependable outcomes from AI models.

Data Quality Monitoring and Maintenance: Continuous monitoring helps to ensure that your data quality remains high over time, preventing the degradation of model performance. Proactive data maintenance, guided by governance practices, leads to sustained accuracy and reliability in AI-driven large data models.

Improved Decision-Making through Transparency: By adopting governance best practices, organizations can clearly document data sources, processing methods, and model decisions. This transparency not only enhances trust in AI-driven models but also enables more informed decision-making by providing insights into the data quality and the factors influencing model outcomes.

Information governance best practices provides a structured approach to data management, leading to enhanced consistency, continuous monitoring, and improved transparency. And, this can not only lower your liability but improve your ability to make the right decisions at the right time!


Rated 0 out of 5 stars.
No ratings yet

Add a rating

Get in Touch

Knowledge Preservation, LLC
567 Woolf Road, Milford, NJ 08848

(973) 494-6068

  • LinkedIn
  • Youtube

Thank you for contacting us. We look forward to connecting with you soon and providing the assistance you require.

bottom of page