When AI Builds AI – Findings from a Workshop on Automation of AI R&D [pdf] (cset.georgetown.edu)

A recent workshop report by the Center for Security and Emerging Technologies (CSET) highlights the increasing use of artificial intelligence (AI) to automate research and development (R&D) in AI itself. The workshop, held in July 2025, brought together experts from AI companies, government, and academia to discuss the opportunities and risks associated with this automation. Key findings emphasize that while advancements in AI R&D automation could accelerate innovation, they also pose significant risks, including diminished human oversight and understanding of AI systems, raising concerns about potential "intelligence explosions" where AI systems might evolve beyond our control. The report identifies critical implications for the AI/ML community, noting that major AI firms are already leveraging AI technologies to enhance their research processes, which could lead to unprecedented acceleration in capabilities. However, the trajectory of this automation remains uncertain, with experts divided on whether it will result in rapid progress or a plateau. The need for better metrics and transparency in AI R&D automation was highlighted, suggesting that systemic collection of indicators could help monitor progress and inform policy decisions to ensure safe and beneficial AI development.