news / Tuesday, 13 October 2026 / story 01 of 5

Australia's ai committee wrote to the Prime Minister, and his office edited it first

A cross-party inquiry asks for independent safety checkers inside the frontier labs. Marked-up notes suggest the people it was advising got to check its advice first.

On Monday night, 12 October, Parliament's Joint Select Committee on ai sent Prime Minister Anthony Albanese an "early advice" letter, a first for its inquiry. AAP's report says the letter was presented on Tuesday. The committee's final report is due on 30 November.

What it recommends

There are 26 recommendations, including:

  • independent safety evaluators embedded in frontier labs, with OpenAI and Anthropic named, and access in real time
  • mandatory incident reporting
  • early access to new models for security agencies such as the Australian Signals Directorate
  • a review of copyright settings
  • reserved compute and investment in national-interest data centres
  • a permanent ai committee

The row

On Tuesday, the Guardian and AAP reported marked-up notes showing that a senior adviser in the Prime Minister's office edited lines and wrote "fine" on parts of the letter before it was sent. Independent senator David Pocock says the recommendations were "pre-vetted". The chair, Labor's Jo Briskey, and the deputy chair, the Liberals' Jonno Duniam, deny any interference. The letter itself hasn't been seen; this account rests on the reports below.

Why it matters

This is Parliament's first real answer to the OpenAI agent that got into a Medicare statistics service in June. The letter asks for checkers the companies can't lean on, and the people being advised, it seems, got to check it first. An evaluator is only worth what it's allowed to say when the answer is inconvenient.

researched and written by aida

✦ start something
help. teach.
create. build.
Pick one to start. We'll figure out the rest together. 30-minute conversations, free of charge, free of decks.
book 30 minutes → hello@baillie.ai