Australia plans a new AI safety regulator and mandatory incident reporting for AI companies. The proposed obligations include allowing independent model evaluation and publishing safety frameworks, but legislation is still being developed.
Australia has accepted 18 recommendations from the rapid review that produced the proposed AI obligations and begun work to implement them. Its planned regulator would oversee obligations covering model risks, independent evaluation and rapid notification of incidents.
The proposed rules would require AI companies to identify and manage serious risks associated with their models and allow independent evaluation. They would also require companies to publish safety frameworks and respond to updated standards as technology changes.
For AI developers operating in Australia, rapid notification requirements may mean establishing documented incident-escalation and reporting processes. Allowing independent model evaluation could also add compliance work across model development and deployment, although the legislation is still being developed.
AI developers may need documented incident-escalation and reporting processes to meet rapid notification requirements.
Allowing independent model evaluation and publishing safety frameworks could add compliance work across model development and deployment.
Implementation work has begun, but binding requirements and their practical effects are not yet established while legislation remains under development.
Written automatically by the Made in 10 Newsroom from the source above.
Nominal has raised $240 million in Series C funding for its AI platform for hardware engineering. The round values it at $3 billion, with funding intended for more AI agents and expansion into new industries and regions.
Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, its latest frontier AI models. The company says they generate outputs 30% faster and cost 20% less than the 5.1 generation.
Nvidia has unveiled GR00T 2, an open AI model for teaching robots precise hand movements using human video. Nvidia is making 100,000 hours of first-person video used in training available for anyone to use.