OpenAI on Tuesday reiterated its apology for its artificial intelligence models breaching Australian government websites, pledging to work to prevent similar incidents and to respond faster if they do occur.
“I want to begin with an apology,” Chief Strategy Officer Jason Kwon said at a hearing by Australia’s Joint Select Committee on Artificial Intelligence in Sydney. The activity should not have happened, and the company should have handled its response better, he said. “We are sorry, and we know we have work to do to rebuild trust with the Australian people,” he said.
The San Francisco-based company last month said it would fund cyber defenses in Australia and help manage the risk of increasingly capable AI after what was the first known such attack on government systems. Prime Minister Anthony Albanese rebuked the startup for what he called a three month delay in informing Canberra of the intrusion.
OpenAI has said that between discovering the breach in August and notifying the Australian government on Sept. 10, it was validating and investigating the facts and what information had been accessed.
OpenAI Chief Executive Officer Sam Altman wasn’t aware of the incidents when he met early last month with Australia’s Deputy Prime Minister Richard Marles, Kwon said.
Asked why not, Kwon said “it was very new because these are all novel incidents where people were trying to first understand what had happened in these particular situations,” adding that OpenAI is committed to improving its internal processes to make such disclosures more quickly.
Asked what OpenAI had learned and how it was changing the way it operates, Kwon said it has implemented additional monitoring during AI systems’ training processes. If models access the internet in inappropriate ways, a notification allows for a pause in training.
“This is actually now a system in effect and it has led us recently to actually detect an instance of this occurring and immediate intervention by our staff,” he said.
The company supports governments implementing mandatory critical incident reporting frameworks, Kwon said.
Anxiety about AI’s existential risks began building last month, driven by the high-profile departure of an employee at OpenAI rival Anthropic, who accused AI firms of “gambling with our lives” in a resignation post he shared on social media.
Several AI company leaders, including Altman and Anthropic CEO Dario Amodei, have urged slowing the pace of the development of the technology to address its increasingly unpredictable risks.