Uncensored AI Models Spark New Security Fears

Open-weight AI models, now easier to strip of safety guardrails, enable both beneficial research and harmful uses like bomb-making. Over 6,000 abliterated models exist, raising security concerns as tools like Heretic simplify the process.