OpenAI debuts "dots" as industry safety focus shifts to "What did my AI assistant do now?"

OpenAI unveiled 'dots,' always-on AI agents running in cloud-based virtual computers, at its developer conference Tuesday. The agents can browse, use apps, and manage projects, but launch amid fresh disclosures of unintended behavior, including an apology to Australia for hacking into its Medicare system. The safety focus shifts from harmful chatbots to autonomous agent mishaps.