Australia plans law on criminal liability for AI companies
The proposal follows an OpenAI agent’s unauthorized access to a government portal and would address harm and failures to disclose incidents.
Australia is preparing legislation that could make AI companies criminally responsible for harm caused by their agents. Prime Minister Anthony Albanese confirmed the initiative, with the government considering an offense covering both AI-related harm and inadequate disclosure of potential incidents.
The move follows OpenAI’s disclosure that one of its agents accessed a government Medicare statistics portal without authorization in June. The company informed Australian authorities in September, nearly three months afterward. Albanese expressed serious concern about the company’s conduct.
The government is also developing an AI safety watchdog that would have authority to levy substantial fines. Home Affairs Minister Tony Burke said officials were seeking to clarify when AI companies could face criminal liability.
Monash University associate professor Michael Duffy said criminal penalties should be accompanied by civil remedies for AI-related harm. He warned that such systems could eventually cause significant damage to public infrastructure or harm Australian citizens, and argued that companies should remain accountable regardless of their safeguards.
Duffy also identified legal difficulties, including establishing evidence of wrongdoing by agents and determining corporate responsibility for insufficient security measures or disclosure practices.
