Hundreds of safety tests for AI models are flawed, say experts

Experts have found weaknesses, some serious, in hundreds of tests used to check the safety and effectiveness of new artificial intelligence models being released into the world.

- Robert Booth UK technology editor

Computer scientists from the government's AI Security Institute, and experts at universities including Stanford, Berkeley and Oxford, examined more than 440 benchmarks that provide a key safety net.

They found flaws that "undermine the validity of the resulting claims", adding that "almost all ... have weaknesses in at least one area". As a consequence, resulting scores might be "irrelevant or even misleading".

Many of the benchmarks are used to evaluate the latest AI models released by big technology companies, said the study's lead author, Andrew Bean, a researcher at the Oxford Internet Institute.

In the absence of nationwide AI regulation in the UK and US, benchmarks are used to check if new AIs are safe, align to human interests and achieve their claimed capabilities in reasoning, maths and coding.