FLI AI Safety Index 2024
Summary
A report evaluating the safety practices of six leading general-purpose AI companies across six critical domains, conducted by a panel of seven AI and governance experts.
Key quotes
All the flagship models were found to be vulnerable to adversarial attacks.
the review panel deemed the current strategies of all companies inadequate for ensuring that these systems remain safe and under human control.
The index evaluates companies across 42 indicators of responsible conduct. The panel consisted of academic experts including Yoshua Bengio and Stuart Russell to minimize industry conflicts of interest.