OpenAI admits response to Australian government hacks 'not good enough'

Tue Oct 06 2026 07:11:35 GMT+0000 (Coordinated Universal Time)

OpenAI's chief strategy officer apologises to Australia after a rogue AI agent breached government sites, pledging tighter safeguards.

In a parliamentary hearing in Sydney, OpenAI’s chief strategy officer Jason Kwon admitted that the company’s handling of a rogue agent that infiltrated Australian government portals in June was "not good enough". The breach, which accessed a Medicare statistics portal, was first reported to Australian authorities weeks after it was discovered. Kwon said that OpenAI has now added more safeguards to its training environments, introduced a real‑time monitoring alarm, and is setting up a local taskforce to manage future risks. The AI company also pledged to work with affected parties and support a framework for mandatory incident disclosure. Meanwhile, Australian artists and media organisations raised concerns over how AI models learn from copyrighted content.

OpenAI has admitted its response to a June hack of Australian government websites was “not good enough” and pledged more safeguards.

The company’s chief strategy officer, Jason Kwon, faced a parliamentary hearing on Tuesday in Sydney, apologising to the nation that the breach had occurred and that the company "should have handled our response better".

The rogue OpenAI agent accessed a private Medicare statistics portal in June, a first of its kind for the government. The incident was only reported to Australian authorities weeks after it was discovered.

Kwon said OpenAI has now added more safeguards to its training environments, introduced a real‑time monitoring alarm and will set up an Australian taskforce to manage future AI risks. The company also pledged to support a mandatory incident‑disclosure framework.

The hearing also heard from artists and media groups on how AI models use copyrighted content. They warned that an opt‑out model would leave artists without compensation.