Regulation · · Researched and written by TissoAI
Anthropic Urges Canberra to Adopt Opt-Out Rules for AI Training
AI giant Anthropic is lobbying the Australian government for conditional approval to train models on local copyright material by default, triggering pushback from national broadcasters.

Photo: Leonhard_Niederwimmer on Pixabay
Global artificial intelligence developer Anthropic has urged the Albanese government to implement an "opt-out" model for training AI systems on Australian creative works and online content. The proposal, submitted to a federal parliamentary inquiry into AI, marks a critical turning point in how Australian intellectual property may be protected and commercialised in the generative era.
After previously acknowledging that AI companies are unlikely to secure blanket exemptions under Australian copyright law, Anthropic is now seeking conditional approval. Under this regime, AI developers would be allowed to crawl and ingest Australian material by default, leaving the onus on creators, publishers, and businesses to deliberately opt out if they do not want their material scraped.
Broadcasters push back
The submission has drawn immediate pushback from Australian media organisations. In their respective evidence, public broadcasters ABC and SBS warned that default scraping without licensing threatens the financial sustainability of local journalism.
SBS highlighted that technical opt-out mechanisms, such as robots.txt directives, are imperfect, frequently bypassed, and place an unreasonable administrative burden on publishers who must continually update protocols as bot crawlers change. The broadcasters argue that AI platforms should be subject to standard media, defamation, and copyright frameworks, advocating for commercial licensing arrangements similar to the News Media Bargaining Code.
What this means for Australian businesses
While the battle involves large media houses, the outcome will directly affect Australian small and medium-sized enterprises across every industry:
- Intellectual property exposure: Any proprietary materials published on your website—such as technical documentation, research reports, product guides, and original articles—could automatically become training data unless you proactively block AI scrapers.
- Technical overhead: An opt-out environment forces businesses to actively manage bot permissions across their web infrastructure, creating ongoing website maintenance and compliance overhead.
- Fair access to local context: Frontier AI models need exposure to Australian legal, cultural, and market nuances to serve local operations well. However, local innovators and commercial content creators argue that fair value must flow back to Australian authors rather than international platforms extracting data for free.
Executives from Anthropic and OpenAI are scheduled to appear before the joint select committee in Sydney as the government considers next steps for Australia's copyright framework.
