Daniel Kang — LLM Hacking Benchmarks

[Navigating Transformative AI] Open Philanthropy recommended a grant of $265,000 over two years to Professor Daniel Kang to support a project to develop benchmarks to measure how well LLM agents perform hacking tasks. This grant was funded via a request for proposals for projects related to benchmarking LLM agents on consequential real-world tasks. This follows our January 2024 support and falls within our focus area of potential risks from advanced artificial intelligence.