CIA Looking Into Uses Cases for Large Language Models, Official Says

Lakshmi Raman, director of artificial intelligence innovation at the CIA and a speaker at a past Potomac Officers Club event, recently said that the agency is in an “experimentation phase” when it comes to determining uses for large language models.

Raman said during a July 14 panel at the Intelligence and National Security Summit that the CIA considers content creation a possible area of application, where the technology could produce drafts that personnel could subsequently edit.

The CIA official went on to note that large language models could also be used to summarize documents involving a high amount of information, Federal News Network reported.

Raman pointed out, however, that her agency’s systems are disconnected from the internet for security reasons. This contrasts with many large language models, which train on online data.

“So that integration aspect also is something that we need to continue to work on and optimize,” Raman said.