AI Surges, Leaving Security Behind: 10,000+ Flaws Found

The rapid advancement of AI in software development has led to an unprecedented surge in code vulnerabilities, overwhelming security teams. Tools like Anthropic's Mythos have identified over 10,000 high- or critical-severity flaws among early partners, with a significant 99% of these issues remaining unpatched as of recent reports.

This rapid discovery contrasts sharply with remediation efforts; fewer than 10% of companies currently fix 90% of identified vulnerabilities within the standard 90-day window. This indicates that while AI accelerates code generation, the capacity for security teams to address the resulting flaws has not kept pace.

Experts emphasize that AI's role in application security is evolving from mere vulnerability detection to crucial remediation at scale. Organizations must integrate AI capabilities with human expertise, focusing on prioritizing exploitable risks rather than solely on the volume of findings. This shift is essential for maintaining confidence in enterprise applications.

Beyond code, the broader AI ecosystem presents new risks. Lack of governance and visibility into AI tool usage, coupled with the expansion of the attack surface, makes it easier for malicious actors to exploit software at machine speed. Addressing these challenges requires clear policies and continuous security integration throughout the development lifecycle.