|
|
|
Photo by conceptcafe/Adobe Stock
|
|
Can artificial intelligence—more specifically, large language models, or LLMs (the technology behind AI chatbots)—be used to launch a large-scale biological attack?
A new RAND study explores this question through a red-team exercise: Teams of experts, role-playing as malicious nonstate actors, were assigned realistic scenarios and tasked with planning a hypothetical biological attack. Some teams had access to the internet, while others had access to the internet and an LLM.
The results show that, while existing LLMs can generate troubling content associated with a biological weapons attack (often mirroring information readily available online), they are not capable of helping to facilitate or plan such an attack.
This suggests that today’s LLMs do not increase the operational risk of a biological attack by a nonstate actor. However, the authors note that ongoing research will be important as LLM technology continues to evolve.
|
|
|
|
|
|
|
Many U.S. veterans struggle with mental health conditions such as depression and PTSD. Good treatments are available, but they don't work for everyone. That's why identifying new, evidence-based treatments is so important. RAND's Rajeev Ramchand recently testified before a House of Representatives subcommittee about one such potential treatment: psychedelics. He discussed the need for adequate funding to develop new treatments, ways to make the research process more efficient, and helping veterans access these treatments if and when they become available. Watch the video »
|
|
|
|
The Department of Defense's Planning, Programming, Budgeting, and Execution System, or PPBE, was established in the 1960s to guide long-term resource development, improve cost-effectiveness, and align national security resources to strategies. Today, as the United States faces a rise in global threats and a rapidly changing technological environment, it's time to reform this system. New RAND research provides insights to help Congress, the Pentagon, and other stakeholders make key updates. Read more »
|
|
|
|
|
Events
|
|
|
|
|
Wednesday, January 31, 2024 – Washington, D.C., and Online
|
|
|
Thursday, February 8, 2024 – Online
|
|
|
|
|
|
