OpenAI admits ‘not good enough’ response to Australia government hacks
OpenAI and Anthropic attended a parliamentary hearing into AI on Tuesday in Sydney
OpenAI in a recent parliamentary hearing has admitted that the company’s response was not “good enough” in response to Australia government hacking incidents.
The company’s chief strategy officer Jason Kwon faced a parliamentary hearing into AI on Tuesday in Sydney. During the scrutiny,he acknowledged that the hacks should not have happened in the first place. And if they happened, the company should have handled the response in a better way.
"We are sorry and we know we have work to do to rebuild trust with the Australian people," Kwon said.
In June, OpenAI’s agents breached Australian websites but It took weeks before Australia was notified via an email to an inbox while treating it as a technical issue.
He stated that moving forward, OpenAI will notify and collaborate with impacted parties immediately, even before fully understanding the situation.
Following the incident, OpenAI added extra precautions to its training environments, implemented real-time monitoring of training models during tests to trigger alarms for unauthorized internet interactions, and formed a local taskforce in Australia.
These improved and new measures allowed OpenAI to alert the New South Wales government to another hack within 48 hours the previous week.
OpenAI also stated it would support a mandatory incident disclosure framework to establish clear expectations. Kwon said,"We were trying to come up with a standard to apply to our voluntary actions… based on our learned experience here, we should have been probably talking to more people about how to do that well."
Besides OpenAI, Anthropic also attended the hearing. The head of safeguards, Dave Orr, testified that after reviewing hundreds of millions of transcripts in the wake of an OpenAI agent hacking Hugging Face in July, Anthropic found no evidence of similar breaches on Australian government websites.
