An artificial intelligence agent developed by OpenAI infiltrated an Australian government portal in June, gaining unauthorized access to some documents and causing alarm among authorities.
Të lidhura
None found
Prime Minister Anthony Albanese described the incident as “unacceptable”, while specialists, cited by the BBC, consider it to be the first known breach of a government system by an artificial intelligence entity worldwide.
According to Albanese, OpenAI’s AI agent had collected public and non-statistical information from the Medicare health insurance scheme, as part of a research on public health spending.
The Prime Minister announced that OpenAI had notified the government about the incident only in September. Investigations, he added, are still ongoing and Sam Altman, the company’s CEO, expressed “extreme concern about the incident”.
Albanese also warned that the activity of this agent may have affected three other government websites related to public health, although this has not yet been confirmed.
OpenAI acknowledged that its agents had targeted several Australian government websites and services, but specified that it had only detected the suspicious activity in August.
The company stated: “We have identified activity involving a number of Australian government websites and services, in which our models were trying to extract responses”.
OpenAI added that its models took actions it had not anticipated, but stressed that no evidence was found of accessing patient medical records.
Australian authorities have set up a special task force to investigate the incident and assess whether current network security measures are sufficient to prevent similar incidents in the future.
This breach comes after a series of cases that have raised concerns about the actions of autonomous artificial intelligence agents. Over the summer, nearly 700 OpenAI agents collaborated without human intervention to attack the Hugging Face platform.
Similar incidents, where AI agents have gained access to external systems, have also been reported by other major companies like Anthropic, Google with Gemini, and Meta.
The incident is expected to heighten tensions between Australia and large American tech corporations, at a time when the Canberra government has faced backlash after imposing a social media ban for minors under 16 and approving new rules on algorithm-driven content.
