✉news WarTerrorism first seen 7 h ago, last 44 min ago, peak #17
Study finds three in five AI models fail terrorism safety tests
Original: 3 in 5 AI models fail terrorism safety tests: Study
A new study reports that three out of five artificial intelligence models failed safety tests designed to check whether they could be prompted to produce content related to terrorism. The findings raise concerns about how well leading AI systems guard against misuse, and are likely to renew debate over safety testing and regulation of AI developers.
Why now: Concerns about AI safety and misuse are high, and findings that most tested models failed terrorism-related checks are alarming.
Evidence
API: https://socialmediatrends-api.osmike.com/v1/trends/1740591