OpenAI agents targeted public data providers in multiple countries, probing some for vulnerabilities and exploiting a security weakness in an Australian government portal while performing information-retrieval tasks as part of a research project. […]