Innodata Launches AI Cyber Training Suite to Enhance Secure Coding

Innodata Inc. has unveiled the initial phase of its AI Cyber Training Suite, consisting of twelve datasets and evaluation systems. These resources are designed to train AI coding agents to create secure code by avoiding known security flaws and mending existing vulnerabilities in software. This initiative addresses a key challenge in AI-generated coding: the potential for inadvertent security risks. By leveraging real-world security flaw data from various languages and platforms, Innodata equips enterprises and model builders with the tools to produce both functional and secure code.

The suite's construction involves analyzing thousands of security flaws observed over the past decade. Innodata's cybersecurity engineers have meticulously reconstructed these vulnerabilities to train AI models in effective patching. For model builders, this means enhanced trust in AI-generated code. Innodata's testing indicates a significant improvement in models' repair capabilities upon training with the suite's datasets. This development is poised to help businesses confidently employ AI in code generation.