This AI Now Institute report examines how the AI industry's energy demands are creating pressure to fast-track nuclear energy development in the US and UK, with three identified risk vectors: weakening of nuclear safety regulation, use of generative AI (LLMs) to expedite nuclear licensing processes, and promotion of unproven advanced nuclear technologies. The authors argue that conflicts of interest arise when AI labs invest in the nuclear providers they rely upon, and that the use of LLMs for high-stakes regulatory documents introduces safety risks that are not yet substantiated. While Australia does not operate civil nuclear power, the report's analysis of AI-driven regulatory erosion and high-stakes LLM deployment in safety-critical domains has relevance for APS practitioners thinking about AI governance boundaries.
Implications
- Monitor APS AI governance teams may want to monitor international developments where AI deployment pressure is undermining established safety regulation, as a cautionary case study for domestic risk tolerance debates.
- Consider Agencies assessing AI use in high-stakes regulatory or licensing contexts could consider the report's critique of LLM-generated compliance documents as a reference when setting appropriate human oversight requirements.
Implications are AI-generated. Starting points, not advice.