Forecasting potential misuses of language models for disinformation campaigns and how to reduce risk

OpenAI · 3 years ago · 48

OpenAI researchers partnered with Georgetown University and Stanford's Internet Observatory to study how language models could be weaponized for spreading disinformation. The project involved a workshop with 30 experts in October 2021 and produced a report analyzing threats and mitigation strategies. The findings provide a framework for identifying and reducing risks from language models being used to amplify false information at scale.