CVE-2026-61439: CVE-2026-61439: Prompt Injection Defense Bypass in PraisonAI InjectionDefense Engine

CVE-2026-61439 exposes a prompt injection defense weakness in the PraisonAI multi-agent framework, allowing attackers to bypass active blocking controls in versions prior to 4.6.78. The InjectionDefense scanner default threshold was set to CRITICAL, permitting high-severity prompts like direct instruction overrides or financial manipulations to pass through.

This flaw enables confidential data compromise and unauthorized agent tool execution. CVE-2026-61439, classified as High severity (CVSS 7.5) and identified as CWE-1188, has a Proof-of-Concept exploit status. To remediate, upgrade to PraisonAI version 4.6.78 or later, configure block_threshold to HIGH, and monitor for HIGH severity logs bypassing blocking.