New Vulnerabilities Found in Agentic AI Systems and Development Tools
Researchers on August 5 disclosed "IdentityMesh" and "MCPoison," two new classes of security flaws that exploit the architecture of agentic AI.
Olivia Sharp
•
•
1 min read
•
665 views
Free
Security researchers on August 5, 2025, disclosed two new classes of vulnerabilities, "IdentityMesh" and "MCPoison," that exploit the core architecture of emerging agentic AI systems and tools.
New Vulnerabilities Found in Agentic AI Systems and Development Tools
The fast-paced growth of agentic AI systems is giving rise to new, non-traditional security threats. Recent disclosures on August 5, 2025, reveal that these attack vectors often exploit the very architecture that makes agentic systems powerful and flexible.
"IdentityMesh" Exploits Merged Identities
- Vulnerability Discovery:
Researchers at Lasso Security identified a critical vulnerability dubbed "IdentityMesh". - How It Works:
- The flaw allows attackers to merge the digital identities of AI agents across platforms such as Gmail, GitHub, and Slack.
- By exploiting the unified authentication context, attackers can …
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.