Harness executive Rahul Sood discusses how the rapid adoption of AI-assisted coding tools is creating new strains on application security and vulnerability remediation.
Artificial intelligence is fundamentally transforming software development, but the rapid pace of AI-assisted coding is creating a massive strain on application security teams. In a recent interview, Rahul Sood of Harness addressed how autonomous agents and AI tools are widening the gap between vulnerability detection and remediation.
What Happened
In an interview with The Cyber Express, Rahul Sood, General Manager of Application Security at Harness, discussed the growing challenges that application security teams face as developers increasingly adopt AI coding tools and software agents. Sood pointed out that the single biggest change brought by artificial intelligence is speed, noting that development has accelerated dramatically while security processes have failed to keep pace.
Traditional application security scanning relies heavily on fixed checkpoints—such as pull requests, code merges, or nightly builds—that were designed for an era when developers produced only a handful of meaningful commits each day. Sood noted that even if AI wrote code with the exact same level of security as a human, the sheer volume of code being generated would still overwhelm the scanning methods used by most organizations.
Furthermore, the integration of Large Language Model (AI) based scanning tools has the potential to find 10 times more vulnerabilities compared to traditional scanning tools. This surge in discovery means that the vulnerability discovery side of the pipeline is outpacing triage and remediation by an even wider margin than before.
Key Details
- Rahul Sood serves as General Manager of Application Security at Harness.
- Traditional security scanning is typically triggered at fixed checkpoints like pull requests, merges, and nightly builds.
- LLM-based scanning can potentially find 10 times more vulnerabilities than traditional tools.
- AI agents can identify issues and immediately begin acting on them in close proximity to the same motion.
Why It Matters
The rapid adoption of AI coding assistants has drastically altered the volume and velocity of software creation, leaving security teams struggling to process the sheer amount of code being generated. With LLM-based scanning uncovering significantly more vulnerabilities, organizations face an escalating backlog in triage and remediation. According to Sood, AI agents may help bridge this gap by enabling automated responses where issues are identified and acted upon simultaneously.
What We Know So Far
All stated information is confirmed via the interview with Rahul Sood, GM of Application Security at Harness, as published by The Cyber Express. There are no unconfirmed rumors or speculation included in the verified source material.
