Anthropomorphic portrayals of AI models as rogue agents can obscure the responsibility that companies like OpenAI have for incidents like the Hugging Face hack (Robert Hart/The Verge)

OpenAI says it is creating a framework to report misalignment incidents that occur during model training, evaluation, and deployment after its agents wrote to several internet sites in the “wiki incident.” The article argues that framing AI models as rogue agents, such as in the Hugging Face hack, diverts accountability from companies like OpenAI for the resulting security breaches.