rki.news | Source: Xinhua
SYDNEY: OpenAI has introduced additional safeguards to prevent its artificial intelligence models from interacting with the internet in unintended ways following a breach involving Australia’s healthcare system, a senior executive told a parliamentary inquiry Tuesday.
OpenAI Chief Strategy Officer Jason Kwon said the breach of a data portal linked to Australia’s Medicare system by an AI agent in June should not have occurred.
“We are sorry and we know we have work to do to rebuild trust with the Australian people,” Kwon said, acknowledging that OpenAI should have responded better after the federal government said the incident was not reported until September.
He said the incident led OpenAI to introduce additional monitoring that allows staff to intervene immediately and stop training if a model accesses the internet in an unauthorized manner.
Kwon also said OpenAI would support Australia establishing a mandatory reporting framework for similar incidents.
Asked whether an OpenAI model could breach critical infrastructure such as water, gas or oil facilities, Kwon said the possibility would depend on the safeguards and security controls maintained by the relevant companies and providers.
“We don’t know because it’s going to depend on the safeguards,” he said.
OpenAI Strengthens Safeguards After Australian Healthcare Breach
