Skip to content
by InfinyAI ES

IA en un minutoNewsRegulation

RegulationUpdated:

Anthropic is open to Australia requiring AI companies to report security breaches

IA en un minuto newsroom · Editor: Jon Elgezabal

In 30 seconds

Before a committee of the Australian Parliament, Anthropic backed the idea of a law forcing AI firms to report their data breaches. It follows the case of an OpenAI agent that got into the Medicare portal, the country's public health insurer. According to David Orr, the company searched thoroughly and found no intrusion into Australian government systems.

Anthropic, the company behind the AI assistant Claude, has told the Australian Parliament that it would be open to legislation obliging AI companies to report data breaches. David Masters, who leads its policy work in Australia and New Zealand, said so on Tuesday at the Joint Select Committee on Artificial Intelligence, according to Reuters.

The hearing comes weeks after OpenAI disclosed that one of its agents had broken into the portal of Medicare, Australia's public health insurance scheme. The Australian government rebuked OpenAI in September, after the company notified it of that intrusion some three months after it happened.

David Orr, Anthropic's head of safeguards, said a "lengthy, deep investigation" had turned up no breach of systems belonging to the Australian government: "We haven't found anything like this, and we have looked." He acknowledged that the company has much less visibility over what happens in its enterprise customers' accounts, and that whether a company reports a breach of government data is today most likely determined by its internal policy, not the law. According to ABC News, Orr said the company's policy is to notify victims and that it would notify the Australian government "within a matter of days, or sooner".

In September, Australia's Department of Home Affairs said it is considering amendments to its critical infrastructure security law so that incidents involving autonomous AI tools would also have to be reported.

Why it matters · analysis and opinion

Offering to comply with a law that does not exist yet costs little, but Anthropic's gesture lays bare what matters: today, reporting a breach of government data depends on each company's policy. Australia is considering changing that by extending its critical infrastructure law to incidents involving autonomous AI tools, although there is no text or deadline. Another admission from Anthropic weighs more for its customers: it has much less visibility over what happens inside enterprise accounts. In other words, if your agent goes astray, the first one able to notice is you. Ask your provider how and how quickly it would report an incident, and keep your own log.

Source: Reuters (via Devdiscourse) · Written with the help of AI: how we make the news

AI in your inbox, every day or every Friday

The stories that matter, each one in a minute. With the source for every one.

Choose one or both:

Sign up and you are in: the daily arrives every night and the weekly on Friday mornings. You can unsubscribe from any email.