2 days ago
- A senior national security official (Hamish Hansford) used AI to write a speech, then blocked FOI requests, demonstrating secrecy and self-preservation.
- This use of AI by a high-ranking bureaucrat constitutes a breach of legal duties under the Public Service Act and PGPA Act, including failure of personal example, care and diligence, and improper use of Commonwealth resources.
- The new APS AI Plan replicates the exact failures of the Robodebt scandal: weak oversight (advisory committee without statutory power), legal vacuum, and voluntary transparency instead of mandatory accountability.
- The plan's timeline prioritizes technology deployment over governance: GovAI goes live in April 2026, but Chief AI Officers are appointed three months later, ensuring no oversight before rollout.
- The plan ignores vulnerable populations, including Indigenous communities (no data sovereignty), women (ignoring warnings of disproportionate job impact), disabled, CALD groups, and those affected by the digital divide.
- Australia's AI strategy surrenders sovereignty to US vendors (Microsoft, OpenAI), with the 'in-house' platform dependent on foreign models and subject to the US CLOUD Act, risking national security.
- Productivity claims are unverified and often fabricated, relying on conflicted industry-funded reports, while evidence shows AI increases workload, causes 'workslop', and adds rework costs.
- Implementation is set for failure: an untrained workforce (only 6% received AI training), integration with legacy systems (e.g., 1980s COBOL), and a history of failed government IT projects.
- The plan represents institutional amnesia, repeating Robodebt's structural failures—weak oversight, legal gaps, and pressure for efficiency over accuracy—with more opaque and powerful black-box systems.