Report: CIA eyes building AI chatbot to rival China
CIA, FBI and friends using AI to uncover threats? What could possibly go wrong?
US spies are reportedly developing their own AI chatbot in a move to top China's prowess.
The US Central Intelligence Agency's Open-Source Enterprise division will soon have a ChatGPT-like large language model (LLM) to provide government snoops with better access to intelligence, according to a Bloomberg report.
The CIA did not respond to The Register's inquiries about the program, including how it will build and train the model, or when it expects it to be ready for use.
Randy Nixon, the director of the CIA's Open-Source Enterprise division, stated that the LLM will allow users — US intelligence agencies including the CIA, FBI, National Security Agency, and military analysts — to see the original source of the information presented and will include a chat feature.
"Then you can take it to the next level and start chatting and asking questions of the machines to give you answers, also sourced," Nixon said. "Our collection can just continue to grow and grow with no limitations other than how much things cost."
Not surprisingly, the AI tooling will reportedly not be made available to policymakers or the public.
American law enforcement and cyber agencies have repeatedly sounded the alarm on allowing China to overtake the US in artificial intelligence. In addition to using AI to find threats, China is building a massive machine learning system to hunt for vulnerabilities, push disinformation campaigns, and develop cyberattacks, according to the FBI, NSA and CISA.
In July, FBI Director Christopher Wray accused China of stealing "more of our personal and corporate data than every nation big or small, combined," and warned that this "information" means the PRC is a "double" threat in terms of AI.
- Uncle Sam mulls spying on clouds being used to train AI
- China identifies AI, optoelectric semiconductors, as challenges it wants to crack
- US cyber chiefs warn AI will help crooks, China develop nastier cyberattacks faster
- Proposed ban on data brokers selling warrantless personal info to Feds revived
China "has already spent years stealing both our innovation and massive troves of data that turns out to be perfect for training machine learning models," Wray said. "And now they're in a position to close the cycle, to use the fruits of their widespread hacking to power with AI even-more-powerful hacking efforts."
Of course, there are also obvious privacy concerns associated with any government agencies amassing a ton of information belonging to its citizens — such as location information from mobile phones, video footage from surveillance cameras, and websites visited via ad trackers.
Currently, data brokers can sell this type of personal information, and more, to law enforcement and federal agencies without a warrant. As such, it's classified as "open source" data.
Nixon seemed to acknowledge these concerns in his interview.
"The scale of how much we collect and what we collect on has grown astronomically over the last 80-plus years, so much so that this could be daunting and at times unusable for our consumers," Nixon said. ®