analiza aiqbench

This study reveals a critical vulnerability in AI coding assistants, which fail to verify software origins despite the availability of trust signals like digital signatures and software bills of materials. The finding that these tools utilize such security data in less than one percent of trials highlights a significant gap between creating provenance information and its actual implementation in automated workflows. For the research community, this underscores the urgent need for developers to either harden AI verification logic or manually audit installations when integrating untrusted code into high-performance or quantum computing environments.