Skip to content

Australia’s DeepSeek ban ‘challenging to enforce’ due to open source code

Australia's ban of the Chinese AI platform on government devices shows the challenge of regulating models which can be freely downloaded, modified and deployed.

James Hennessy
Make us a preferred source
The Labor government has banned the use of DeepSeek's AI models on government devices, but the open source nature of the platform creates a complex challenge. Sheldon Cooper / SOPA Images.

The decision to ban Chinese AI platform DeepSeek from Australian government-issued devices and systems has raised questions about the complexity of enforcing national security regulations and restrictions on open source AI.

DeepSeek shocked the artificial intelligence industry and rattled tech markets in January when it released a new "reasoning" model, R1, which demonstrated capabilities rivalling bleeding edge models from well-resourced US counterparts like OpenAI, Google and Anthropic. It also showed significant improvements in efficiency over those competitors, making it more economical to run for users and developers.

Unlike models such as OpenAI’s GPT-4 — but like Meta’s Llama — DeepSeek’s models are open source, with its code and underlying model ‘weights’ freely available under permissive licences. This allows developers to download, modify and repurpose them, provided they adhere to the associated licensing terms.

But this could make it far more difficult for the government to enforce a total ban as it has with ByteDance's TikTok.