How law-abiding is AI? ETH researchers put it to the test
Briefly

The researchers have provided the first detailed technical interpretation of the EU AI Act for General Purpose AI models, translating legal requirements into measurable standards.
Their practical approach allows model developers to evaluate their alignment with future EU legal requirements through runnable benchmarks that have not existed before.
Testing involved twelve popular generative AI models like ChatGPT; the translation of regulatory requirements into practical benchmarks is crucial for ensuring ethical AI use.
The EU AI Act aims to enhance public trust and reduce the risks associated with AI technologies by demanding transparency, fairness, and accountability in AI models.
Read at ETH Zurich
[
|
]