COMPL-AI Framework Finds OpenAI, Meta AI Models Fail EU Act Standards in Cybersecurity, Bias

Oct 16, 2024, 08:32 AM

LatticeFlow, in collaboration with ETH Zurich and INSAIT, has launched COMPL-AI, the first open-sourced EU AI Act compliance evaluation framework for generative AI models. According to an exclusive report by Reuters, the framework assesses AI models from major companies including OpenAI, Meta, Alibaba, Anthropic, and Mistral against the forthcoming European regulations. The evaluation revealed that while these models scored well in certain areas, they fall short in key compliance aspects such as cybersecurity resilience and discriminatory outputs. The findings highlight significant compliance gaps and underscore the urgency for Big Tech to address issues related to transparency, cybersecurity, and bias as they race to meet the new standards set by the EU AI Act.