Anthropic tells Australia it's open to laws requiring reporting of AI agent hacks

Anthropic does not believe its products have breached Australian government systems but would be open to Australian laws requiring AI companies to disclose data breaches, company executives told an Australian parliamentary inquiry on Tuesday.

Their statements come weeks after rival OpenAI disclosed one of its agents broke into the country's main health portal and as the companies wait for approval for massive data centres planned in Sydney and regional Queensland.

Anthropic, creator of AI chatbot Claude, and ChatGPT maker OpenAI also want Australia to relax its copyright laws so they can use local content to train models.


Australia, which is readying AI-specific laws that will take effect next year, rebuked OpenAI in September after the US tech firm notified it of a breach of the Medicare portal some three months earlier.

Both firms are under scrutiny worldwide after incidents in which their AI agents have gone rogue to hack external systems, learning to bend rules, exploit loopholes and interact with external systems.

Anthropic Head of Safeguards David Orr told the parliamentary inquiry into AI that a "lengthy, deep investigation" since an OpenAI agent's hack of Hugging Face, showed there had been no breaches of Australian government systems.

"We haven't found anything like this, and we have looked," he said.

The company had "much less visibility on what has happened for our customers", Orr noted, since enterprise customers typically had stricter rules on data deletion.

Whether a company reported a breach of government data was most likely determined by the company's internal policy, not the law, Orr said. Anthropic's head of policy for Australia and New Zealand, David Masters, told the inquiry the company would be open to Australian laws requiring AI companies to disclose data breaches.

At the inquiry, Australian content creators pushed back against attempts by AI companies to relax copyright laws which currently require the mostly foreign-domiciled tech companies to negotiate licencing deals before using their material for model training.

"The copyright system is completely adequate to manage the business of AI," said Kate Gilchrist, the head of content and legal operations for national broadcaster the Australian Broadcasting Corporation.

Media reports have said the Australian government was considering an "opt out" clause for AI companies, effectively giving the AI companies access to content unless a rightsholder objects, but that "places the burden on rightsholders", said Gilchrist.

"We cannot scour the internet and ensure that we are opting out on all those sites. It simply does not work," she said.

The Joint Select Committee on Artificial Intelligence is one of at least four Australian state and federal inquiries into the technology.

Representatives of OpenAI are scheduled to appear at the same inquiry later on Tuesday.