Weaponizing ChatGPT? National Security and the Perils of AI-Generated Texts in Democratic Societies

The proliferation of Artificial Intelligence-generated texts creates enormous opportunities for the economy and society, but also introduces potential risks. Can language models like ChatGPT be weaponized in ways that have adverse implications for national security, including elections and battlefield command and control? Does defense have any advantages in a world of large-scale AI-generated misinformation? Are there potential upsides of AI-generated text in terms of democratic governance or national security? In this talk, Professor Kreps brings to bear a range of original experimental evidence to suggest that the answers to all of these questions might be affirmative and offers guidance on how to harness the prospects and guard against the perils of generative AI.