Anthropic\u2019s special envoy, the former US ambassador to Australia Jeffrey Bleich, told the committee that Anthropic had \u201cnever tried to dictate to Australia\u201d on copyright.<\/p>\n
\u201cWe\u2019ve just shared information about how the training works, and then discussed what its implications would be with respect to our capacity to operate here,\u201d he said.<\/p>\n
Anthropic suggests current Australian laws mean AI companies need to obtain a licence for the use of every piece of content on the internet as training material. Bleich said this was \u201ctechnically impossible\u201d.<\/p>\n
Anthropic was keen to explore what the Australian government wants to do to incentivise training in Australia, and Bleich said Australia could shape frontier models if they were trained here, otherwise it would ultimately rely on models trained elsewhere.<\/p>\n
Sign up for the Breaking News Australia email<\/a><\/p>\n OpenAI apologised for its AI agents attacking a number of Australian government websites in June<\/a>. The multibillion-dollar company notified the government via an email to a public-facing address<\/a> three months after the breach.<\/p>\n \u201cI want to begin with an apology. During internal training and evaluation, our models accessed Australian government websites in ways they were not directed to. That should not have happened. We also should have handled our response better,\u201d OpenAI\u2019s chief strategy officer, Jason Kwon, told the committee on Tuesday afternoon<\/a>.<\/p>\n \u201cWe are sorry, and we know we have work to do to rebuild trust with the Australian people. We are committed to doing that work.\u201d<\/p>\n Following the revelation of a separate hack on the NSW national parks and wildlife service last week<\/a>, Kwon told the committee OpenAI was reviewing logs of AI agent training dating back to November 2025, which identified the latest hack activity taking place in June.<\/p>\n He promised that should more incidents be identified, government agencies would be notified \u201cvery, very quickly\u201d.<\/p>\n OpenAI representatives, while apologetic, downplayed the agent activity on Australian government websites as not \u201cnot super sophisticated\u201d compared to the Hugging Face attack<\/a>. Independent senator David Pocock questioned why the company didn\u2019t contact ministers and instead sent an email \u201cto some arbitrary department email\u201d after the Medicare incident.<\/p>\n \u201cI think that on retrospect, we should have done what you were suggesting \u2026 I think people were thinking about this as a technical situation, and they wanted to contact the technical counterparties \u2013 but it\u2019s not good enough,\u201d Kwan said.<\/p>\n Anthropic\u2019s head of safeguards, Dave Orr, told the committee the company had reviewed hundreds of millions of transcripts of its models and had not found any cases in which it interacted with Australian government systems in an unauthorised way.<\/p>\n He could not say whether Anthropic\u2019s customers may have used the models on Australian government systems, given the company\u2019s \u201czero data retention\u201d policy.<\/p>\n
![\"\"](\"https:\/\/www.newsbeep.com\/il\/wp-content\/uploads\/2026\/10\/1791265211_714_2998.jpg\")
<\/p>\n![\"Anthropic](\"https:\/\/www.newsbeep.com\/il\/wp-content\/uploads\/2026\/10\/1484.jpg\")
Anthropic says AI agents didn\u2019t breach Australian government websites \u2013 video\u2018Artists will become roadkill\u2019<\/p>\n