CIA Director of Artificial Intelligence, Lakshmi Raman says that AI is making US adversaries more dangerous but also presents the agency with an important resource.
In an interview with Politico, Raman said that the CIA uses AI in much the same way that average users do, automating low-level office tasks like filtering and summarizing data or doing human language translation.
When asked about how US adversaries were using AI, Raman said that the CIA was concerned about “things like deepfakes and disinformation as well as cybersecurity risks, how is AI being used to create more phishing emails, or generating more malware.”
On the AI risk presented by China, Raman said the CIA had a very different approach to AI development. She said the CIA was concerned with “responsibility, security, safety, ethics,” in its approach to employing AI.
Raman said that China’s AI capabilities were “growing every which way” and were not being developed according to “our democratic norms.”
CIA’s version of ChatGPT coming soon
Earlier this week Bloomberg revealed that the CIA was building an AI chatbot similar to ChatGPT to help its agents to sift through the mountains of data that the agency collects.
Randy Nixon, the CIA’s director of Open Source Enterprise told Bloomberg that the flood of “public data” made it increasingly difficult for its agents to find relevant data.
Nixon didn’t define what the agency felt constituted “public data” but said that once they have their AI tool in place the CIA’s data collection “can just continue to grow and grow with no limitations other than how much things cost.”
Nixon did say that the tool would follow US privacy laws but the CIA has a less-than-clean record as far as that’s concerned. The potential involvement of the CIA with another R. Nixon in the Watergate scandal is an ironic case in point.
More recently the CIA has reportedly bought commercially available cellphone location data to avoid having to obtain the necessary warrants. AI has already thrown up a lot of red flags around surveillance and privacy issues.
Creating a reliable AI tool that ticks all the legislative boxes and adds value to the CIA will be a tough challenge for the agency.
Raman highlighted the lack of available skilled engineers, saying that the CIA’s “demand far exceeds the supply.” As a result, the agency is heavily reliant on private enterprise.
She didn’t mention which AI tech companies the CIA was working with or which model would be used to create its AI tool.
