Automation of AI research and development poses significant implications, but the extent of its impact remains unclear. Empirical data is necessary to understand the effects of AI R&D automation, but current benchmarks may not accurately reflect real-world automation or its broader consequences. The acceleration of capabilities versus safety progress is a key concern, as it may have far-reaching repercussions. State-aligned threat activity elevates the stakes, shifting the focus from criminal to geopolitical implications that extend beyond the immediate target. The lack of comprehensive data hinders the ability to assess the true impact of AI R&D automation, making it challenging to develop effective strategies to mitigate potential risks1. This uncertainty matters to practitioners, as it underscores the need for more nuanced and informed approaches to AI R&D automation, lest they inadvertently exacerbate existing security vulnerabilities.