Anthropic declined an Australian Senate AI hearing as officials investigate an OpenAI agent breach and consider mandatory AI incident reporting.