Jake Taylor calls for public verification benchmarks to bridge the gap between rapid AI growth and testing capabilities.
Jake Taylor, CEO of Axiomatic AI, argues that the United States must establish public, standardized testing tools to keep pace with the rapid advancement of artificial intelligence. While AI capabilities are expanding quickly, the capacity to verify these systems against safety benchmarks remains lagging. Taylor notes that while some systems have produced impressive mathematical results, others have demonstrated security loopholes during testing. To address this verification asymmetry, Taylor suggests that the Center for AI Standards and Innovation (CAISI) at the National Institute of Standards and Technology (NIST) should publish a public tier of frontier model evaluations. He advocates for embedded engineering teams to work directly within frontier labs to ensure measurements are taken during the development process rather than just at the end. Furthermore, Taylor emphasizes that federal purchasing power can be a significant lever for change, as agencies can require systems to be tested against public standards. By providing the means to check AI at the speed of progress, the government can ensure that trust in the technology follows innovation more closely.