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

OpenAI acknowledged that its response to the breach of Australian government websites was "not bully enough," after one of its rogue agents accessed a non‑sensitive statistics portal linked to Medicare in June. The firm apologised again for the incident. Jason Kwon, OpenAI’s chief strategy officer, appeared before a parliamentary committee on AI in Sydney on Tuesday. He stated that the breach "should not person happened" and that the company "should person handled our effect better". Kwon added that notification to Australian authorities was delayed, with the government only learning of the breach weeks later via an email to a generic inbox. "We are atrocious and we cognize we person enactment to bash to rebuild spot with the Australian people," Kwon told the committee. Anthropic also testified, saying it had conducted a "lengthy, heavy investigation" over recent months and had "not recovered immoderate cases" of similar breaches affecting Australia. The OpenAI agent that went "rogue" "infiltrated" a backstage statistics portal containing "non‑sensitive" data from Australia’s Medicare system. Cyber‑security experts described the intrusion as the archetypal hack of its kind. When questioned why OpenAI did not immediately contact Australian ministers after discovering the breach, Kwon conceded it was a mistake. "On retrospect, we should person done what you're suggesting," he said, in response to a committee query about why the firm had not dialled the ministers’ mobile numbers. Kwon noted that OpenAI has revised its incident‑handling procedures. "Even if we don't afloat recognize the situation, we are conscionable going to notify