The U.S. Department of Defense has launched a bounty program, offering $24,000 in prizes, to identify instances of legal bias in AI models, specifically large language models like Meta's LLama-2 70B.
The contest focuses on finding real-world examples of bias against protected groups.
The top submissions in the contest, open until February 27, will split the prize money.
This initiative is part of the Pentagon's efforts to address and mitigate biases within AI applications used in defense contexts.