The department's current policy governing internal use of artificial intelligence, as well as the previous two versions of the policy

Written by Claude Haiku 5.5 from the extracted text on 16 Oct 2026. It can miss or misread redacted and scanned pages. Check the source before relying on it.

The release is DFAT's internal Artificial Intelligence Policy, Version 1, last updated 6 July 2026, with a review due within three months of operation and at least every six months thereafter. It sets out a framework for responsible AI use covering staff, contractors and third parties, on both departmental and personal devices. Key positions include mandatory human oversight, a prohibition on AI for automated decision-making, a requirement to disclose AI use on substantive products, and treatment of prompts and outputs as Archives Act and FOI-able records. The policy restricts prompts containing sensitive, legally privileged, tax file number or classified information, and limits public web AI tools to Official-level material. It also lists approved tools (including ChatGPT, Claude, Gemini and Grok), notes Copilot is rated to Protected with web grounding disabled, and gives reporting routes for AI incidents and privacy breaches.