Government Outsources AI Security Law to the AI It's Meant to Regulate
After OpenAI urged Canberra to tighten the rules around artificial intelligence, officials asked a chatbot to write those rules — and then to check its own homework.
Responding to OpenAI's call for clearer, better-coordinated AI security laws, the federal government has announced that the actual text of its new legislation will be drafted by an AI system built by OpenAI. Officials describe the arrangement as efficient, noting that the same model can review its own draft for compliance immediately after finishing it, with no separate reviewer required.
The model's self-assessment reportedly concluded that its draft law was fair, clear, and well within the bounds of its own judgment, recommending only a minor amendment expanding its own authority to interpret future amendments.
"We think this represents genuine multi-stakeholder collaboration," said a spokesperson for the newly formed AI Security Steering Committee, a body whose three members are the model, a printout of the model's previous answer, and a chair currently unfilled pending the model's recommendation on who should fill it.
A separate briefing note confirms the Prime Minister will still travel to the US to meet Apple's Tim Cook, where the two are expected to discuss whether the phone used to read the new AI law aloud in parliament should also be asked to summarize its own settings for the public record.
The inquiry itself remains ongoing, with its next scheduled task being to ask the same chatbot whether the inquiry should continue.