CYBERSECURITYTRACKER
TRACKING6,506 stories in this site build1,309 vulnerability news stories in this site build
Permanent story citation

The best defense against AI attacks turns out to be a skeptical human

This page keeps the story as Cybersecurity Tracker first published it. If the tracker later corrects it, the correction appears below the original and never replaces it.

Back to newsStory 2458

As cited

Copy frozen at (site build).

ai security

The best defense against AI attacks turns out to be a skeptical human

A SANS survey of 536 IT and security professionals found that 78% of practitioners now use generative AI in daily security work, up from 50% the previous year. However, reliability has not kept pace with adoption, with 63% reporting significant shortcomings in AI detection and response capabilities.

Why it matters: Security teams relying on AI for log analysis and incident response need to maintain skeptical oversight and validation, as AI tools show substantial gaps in accuracy and reliability that could miss genuine threats or produce poor guidance.

Source published
First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

ai security

The best defense against AI attacks turns out to be a skeptical human

No summary had been written when this copy was frozen.

First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

ai security

The best defense against AI attacks turns out to be a skeptical human

Seventy-eight percent of security practitioners now use generative artificial intelligence (AI) in daily work, according to the 2026 SANS AI Survey of 536 IT and security professionals. Adoption nearly doubled from the prior year, though reliability concerns persist, with 63 percent reporting significant shortcomings in AI detection and response capabilities. Human skepticism remains critical to managing AI limitations in security operations.

Why it matters: Security teams relying on AI for log analysis and incident response must validate AI outputs independently; widespread adoption without verification creates blind spots in threat detection and response workflows.

Source published
First seen by Cybersecurity Tracker

Source attribution

Glossary