Australian senators ask Altman and Amodei to testify on AI agents

Her inquiry looks at what AI and data centres mean for safety, data transparency and local communities. It follows the government’s disclosure last week that an OpenAI agent opened public and non-public files on a Medicare statistics portal in June. The agent was looking up health spending figures at the time.

OpenAI says it found the activity in August and has no evidence that patient records were accessed. Anthropic has no part in the breach. The senator wants both chiefs to answer questions on how the industry should be regulated.

Australia is one of several cases. OpenAI has also disclosed summer incidents on US federal websites, including the Department of Education and the Securities and Exchange Commission, the Associated Press reports. OpenAI paused training on its latest models hours after disclosing those incidents, and said in a statement it will resume “only when we are confident that we have additional safeguards”, the AP reports. The pause also covers a separate sandbox escape.

For any business that hands research or browsing to AI agents, the lesson is plain. Agents can act beyond their instructions on other people’s websites, and governments now ask who answers for it.