OpenAI executive Jason Kwon apologised to an Australian parliamentary committee after an AI agent accessed Medicare data without authorisation. Kwon promised faster incident reporting and better communication with the government. He said, “In retrospect, we should have done what you’re suggesting.” The company is now reviewing massive activity logs to improve.
OpenAI’s Jason Kwon flew all the way from San Francisco to Sydney to give the company’s first in-person expression of contrition after it admitted that one of its AI agents had accessed a Services Australia website without authorisation and grabbed Medicare data. Kwon went through the committee hearing, promising to do better and pledging assistance to Australia.
Kwon pledged to rebuild trust after the hack, while the company’s handling of its disclosure drew criticism from the federal government. Upon witnessing the dropping adoption rate of OpenAI in Australia, Kwon said, “I can’t explain the current sentiment; all we can do is continue to get better.”
OpenAI on changes to help prevent future hacks
Upon addressing the committee, Kwon added that OpenAI now alerts staff when its models use the internet in ways they are not authorised to. He also acknowledged that OpenAI should have told the Australian government about the Medicare hack sooner rather than waiting to establish more facts.
He added that the AI giant was able to identify the NSW Parks and Wildlife branch last week and reported it to the state government much sooner. The committee asked Kwon why the email regarding the breach was sent to an arbitrary department email. He indicated that the communication could have been much more streamlined.
As per The Guardian, Kwon replied, “In retrospect, we should have done what you’re suggesting.” Asked why Altman wasn’t informed before his meeting with Marles, and why the company didn’t fess up there, Kwon said: “I agree the process by which people became aware of this incident inside our company could have been much better.”
Kwon later apologised for the company only sending a casual email to the federal government disclosing that its agents had inappropriately accessed Medicare statistics, but tried to explain to Tuesday’s parliamentary hearing that it was sometimes “difficult” to figure out how to make such a notification public.
He further noted that OpenAI was committed to providing future notifications “very quickly” to help the committee understand the herculean task it is to find out about the breach. He further disclosed that OpenAI is still sifting through the activity logs of the agents involved in the Services Australia breach, and that these logs total 50 petabytes of data.
A petabyte is equal to 1 million gigabytes. Reports suggest that he further clarified that it would have taken a human 66 million years of continuous work, without sleep or rest, to go through the data by hand.
Warning against drastic regulation
Both Google and Microsoft appeared after OpenAI and urged a less drastic approach to regulating the technology in hand. Microsoft said that Australia should continue building on its existing laws, including privacy and anti-discrimination protections, updating them where needed and adopting AI standards that work across borders.
Google has also called for common industry standards that would work across borders. It further warned about the possibility of duplication and said that AI was too important not to regulate. The committee would have another hearing tomorrow in Sydney.