Intelligence officials across the Federal government are raising concerns about the escalating challenges and implications of artificial intelligence (AI), particularly regarding data authenticity.
During day one of the 2024 AFCEA/INSA Intelligence & National Security Summit on Aug. 27, several officials underscored the essential need for human oversight in AI-generated data and called for the development of strong ethical frameworks to address the limitations of current AI systems.
“One of the really interesting things about the last few years has been the rise of generative AI, and the way that’s interacted with [how] we try to make sense of information,” said Wil Corvey, program manager for the Information Innovation Office at the Defense Advanced Research Projects Agency (DARPA).
He explained that the problem with many uses of AI is the assumption that “[it] could produce globally authentic data,” but in some cases it’s not succeeding in that goal. Corvey explained that users of AI, particularly researchers, must “unpack AI generated information with two key things in mind; authenticity … and techniques before we utilize this information,” Corvey said.
Mark Munsell, director of data and digital innovation at the National Geospatial-Intelligence Agency, echoed Corvey’s comments, adding that to combat the data authenticity challenge, “AI systems must be trained effectively to ensure accurate data is produced.”
“We must take extra steps to ensure that the artificial intelligence producing this data is completely accurate. For us, the stakes are high, especially when identifying a potential kinetic target. We need robust frameworks and an additional level of responsibility to guarantee the accuracy of the information,” Munsell said.
“The safer and higher quality we make AI, the more likely we are to make faster and more accurate human-assisted decisions,” he said, emphasizing that human judgment remains a critical piece of that puzzle.
Jason Richards, deputy assistant director for the Directorate of Intelligence at the FBI, similarly emphasized that while AI can help identify targets, “human judgment is crucial for final determinations.”
