KORA: A public benchmark for AI Child Safety across frontier models (korabench.ai)

KORA has launched a public benchmark focused on AI Child Safety, evaluating various frontier models such as Anthropic's Claude and OpenAI's GPT series. This benchmark provides up-to-date scores, historical data, and trends, allowing researchers and developers to audit and run tests independently using open-source code. The leaderboard ranks models based on their performance in ensuring child safety, with Claude Haiku 4.5 leading the way at 76%, followed closely by OpenAI's GPT-5.2 at 75%. This initiative is significant for the AI/ML community as it directly addresses the pressing need for responsible AI development, particularly when it comes to protecting vulnerable users such as children. By standardizing measures of safety across different AI models, KORA not only fosters greater accountability among AI developers but also encourages improvements in safer AI design. The methodology and transparency of results may also pave the way for regulatory discussions and public trust in AI technologies, making child safety a priority in ongoing AI advancements.