Data Experts Highlight Need to Streamline Data to Secure GenAI Applications

Government experts said streamlining data is the key to the secure adoption of generative artificial intelligence for various federal applications.

Speaking at an ATARC panel Thursday, Chakib Chraibi, the chief data scientist at the Department of Commerce’s National Technical Information Service, said unstructured data brings cybersecurity challenges, including data poisoning, hallucination and prompt injection that cause harmful or biased AI outputs.

According to Chairabi, data scientists must work with subject matter experts to ensure the federal government uses complete, consistent and unbiased data in their GenAI applications, Nextgov/FCW reported.

The National Institutes of Health and the Chief Digital and Artificial Intelligence Office are using AI software to convert unstructured data into searchable information to support various initiatives. 

The National Science Foundation, meanwhile, is planning to deploy a chatbot that uses large language models to perform retrieval or predictive analytics to improve customer experience.

Terry Carpenter, chief information officer at the NSF, highlighted the need to clean up data for GenAI applications, noting that unstructured historical government data can impact how an AI algorithm works.