US and Australia confront AI regulation challenges

Washington officials say open‑weight Chinese AI models, which can be downloaded and self‑hosted, are impossible to ban once they are in enterprise servers. The models now represent roughly 45‑46% of enterprise API tokens on OpenRouter, and companies such as Coinbase, Uber, Cursor and Airbnb have adopted them to lower AI costs. After a State Department warning on July 8, congressional committees expanded investigations and the U.S. is shifting from outright bans to procurement restrictions, compliance requirements and security audits.

In Australia, the absence of a workable copyright framework is blocking AI investment. Rights‑holder bodies have rejected a proposal that would offer a text‑and‑data‑mining exemption in exchange for payments into a creators’ fund and billions in data‑centre investment. The recent Anthropic settlement in the United States, which highlighted the risk of training on copyrighted works, underscores the legal uncertainty. Australian policymakers face pressure to resolve the issue so AI development can proceed without exposing firms to massive liability.