Open-source AI models easily manipulated to generate antisemitic content

Study finds alarming gaps in safety guardrails across popular open-source Large Language Models, new ADL research shows.