MEDIUMVulnerability
Global

Attackers Could Exploit AI Vision Models Using Imperceptible Image Changes

·Source: SecurityWeek

Updated:

Executive Summary

Cisco’s AI security researchers have analyzed ways to target vision-language models (VLMs) using pixel-level perturbation. The post Attackers Could Exploit AI Vision Models Using Imperceptible Image Changes appeared first on SecurityWeek .

Analysis

Cisco’s AI security researchers have analyzed ways to target vision-language models (VLMs) using pixel-level perturbation. The post Attackers Could Exploit AI Vision Models Using Imperceptible Image Changes appeared first on SecurityWeek .
Source Attribution

Originally published by SecurityWeek on May 7, 2026.

Related Threats

MEDIUMVulnerability

Go-Ahead for AI Chip Sales to 10 Chinese Firms Raise Alarms

<img src="https://ismg-cdn.nyc3.cdn.digitaloceanspaces.com/articles/go-ahead-for-ai-chip-sales-to-10-chinese-firms-raise-alarms-image_small-9-a-31697.jpg" align=right hspace=4><b>Reports: Trump Administration Approval of Nvidia H200 Sales Poses Frontier AI Risks</b><br>Trump administration discussions on AI governance with China are colliding with reports that Washington may permit expanded Nvidia

Bank Info Security
MEDIUMVulnerability

Taiwan Incident Highlights Cybersecurity Gaps in Rail Systems

A Taiwanese student experimenting with software-defined radio technology shut down three bullet trains for nearly an hour, leading to an anti-terrorism response.

Dark Reading
MEDIUMVulnerability

No need to hack when it&#8217;s leaking: Dalbir Singh &#038; Associates law firm edition

Dalbir Singh &#38; Associates ignored multiple attempts at responsible disclosure but finally locked down its misconfigured Amazon bucket, only to expose it again. Now the data is in the hands of criminals trying to extort them. On April 6, DataBreaches reported on a misconfigured Amazon bucket belonging to an immigration law firm in New York.... Source

DataBreaches.net