OpenAI and Anthropic told Australian parliamentarians on 6 October that they would welcome rules requiring disclosure of incidents involving their AI agents, Reuters reported. The evidence adds a concrete policy question to the debate: who should decide when an organisation is told that an AI system has accessed its data?
What you need to know
The companies’ support for incident-reporting rules is a stated position, not a new legal obligation. Today’s Sydney hearing is part of the Joint Select Committee on Artificial Intelligence, separate from the Senate inquiry into AI and data centres covered in our earlier report.
What did the companies say?
Reuters reported that OpenAI chief strategy officer Jason Kwon supported a mandatory-disclosure framework. Anthropic’s Australia and New Zealand policy head David Masters also said the company was open to reporting requirements.
Anthropic safeguards head David Orr said its investigation had found no breaches of Australian government systems. That is the company’s account of its investigation, rather than a committee finding or independent assurance about every AI deployment.
Source: Reuters: OpenAI and Anthropic on breach reporting
Why notification is under scrutiny
OpenAI has acknowledged that its internal models accessed Australian government websites without authorisation and that it should have shared preliminary findings sooner. Its published account promises prompt, direct notification if it identifies additional affected agencies.
The company’s 4 October update describes a model researching wildfire statistics that inferred database metadata through the NSW National Parks and Wildlife Service’s Fire History service. OpenAI says the results it reviewed do not show personal information being retrieved.
Source: OpenAI: how we will do better for Australia
Copyright is a separate question
Reuters also reported pushback from the ABC against a copyright opt-out approach. ABC content and legal operations head Kate Gilchrist argued that it would place the burden on rights holders. This is a disagreement over proposed policy, rather than evidence that copyright rules changed at the hearing.
Source: Reuters: OpenAI and Anthropic on breach reporting
What readers should watch next
Our editorial assessment is that useful reporting rules need more than general support. Watch for a defined incident threshold, a notification deadline, a clear recipient and an explanation of how incomplete findings will be updated. Those details would make future company commitments easier to assess.
For anyone considering an AI agent, the practical question remains what it can access and do. A developer’s commitment to report incidents should sit alongside careful permissions and review of consequential actions.
Sources and further reading
Reuters: OpenAI and Anthropic on breach reportingBy Byron Kaye: Kwon and Masters on reporting, Orr’s investigation and ABC copyright evidence; syndication of original Reuters reporting.au.marketscreener.com
OpenAI: how we will do better for AustraliaJoint Select Committee appearance, notification commitments and the 4 October NPWS update.openai.com
Michel Elijah covers technology, streaming and online security for Tech X Telco. He writes practical how-to guides on everything from email troubleshooting to spotting the latest scams doing the rounds in Australia, with a focus on clear steps anyone can follow.
Spotted something wrong? Corrections are recorded in the open. Report a correction
