AI
Anthropic gives Australia direct AI adoption data feed in first National AI Plan MOU
The Economic Index data-sharing clause gives Canberra sectoral AI adoption telemetry no comparable jurisdiction holds — Australian deployers gain a reference point, not a compliance shortcut, and cross-border obligations remain unaffected.
2026-04-28 · 2 min
Australia and Anthropic sign MOU with direct government data feed under National AI Plan
Anthropic's Economic Index — sectoral adoption telemetry shared directly with a federal government — is the most novel element of a Memorandum of Understanding signed in Canberra on DATE TK between the Australian government and the frontier-model developer, the first arrangement registered under the National AI Plan launched in late 2025.
The Memorandum of Understandinghttps://www.industry.gov.au/publications/memorandum-understanding-between-australian-government-and-anthropic-collaboration-ai-opportunities bundles four operative commitments: joint evaluations with the Australian AI Safety Institute, Economic Index data sharing on AI adoption across natural resources, agriculture, healthcare and financial services, A$3M in research partnerships, and a Sydney office opening in 2026. Signed by Anthropic CEO Dario Amodei and Prime Minister Anthony Albanese, it is the first arrangement registered under the National AI Plan.
The Economic Index clause gives Treasury and the Department of Industry a data feed no comparable jurisdiction currently receives at this granularity.
The A$3M research tranche names four recipients: Australian National University, Murdoch Children's Research Institute, the Garvan Institute of Medical Research and Curtin University, working on clinical genomics, precision medicine, paediatric research and computing education using Claude.
"Australia's investment in AI safety makes it a natural partner for responsible AI development. This MOU gives our collaboration a formal foundation," Amodei said.
What it does, and does not, change for deployers
The MOU operates at the foundation-model provider layer. Joint evaluation work with the AI Safety Institute covers model-layer risks — capability evaluations, security testing, frontier-model behaviour. A health-tech firm in Melbourne running Claude inside a clinical-decision support tool does not inherit an assurance artefact from that work. Therapeutic Goods Administration obligations, Privacy Act 1988 duties under the Australian Privacy Principles, and sector-specific duties of care still attach to the deployer independently of what Anthropic and the AI Safety Institute jointly publish.
Joint evaluation outputs become a reference point Australian deployers can cite in their own risk documentation, in much the way NIST AI RMF mappings function in US procurement. That is supporting evidence, not a conformity shortcut.
Cross-jurisdiction read
The MOU does not create GPAI passporting. Anthropic's obligations under the EU AI Act's general-purpose AI provisions — which apply from August 2026 for the high-risk deployment surface — run on a separate track. Nothing in a bilateral Australian instrument substitutes for the EU AI Office's GPAI designation process or the model-card disclosure calibration the AI Office is finalising. UK AI Safety Institute evaluations operate under a separate voluntary protocol. Australian deployers selling into those markets still face the originating jurisdiction's framework.
The data-centre and renewable-energy alignment angle sits outside the model-governance layer and is flagged for the regional infrastructure desk.
The first joint evaluation output from the Australian AI Safety Institute, the Sydney office opening in 2026, and whether the National AI Plan's second arrangement — expected to follow with another frontier-model provider — replicates the Economic Index data-sharing clause or treats it as Anthropic-specific remain the milestones to track.