// Security Advisory Triage
PraisonAI prompt injection defense bypass patched in 4.6.78
GitHub advisories — pip · published 2026-10-08 13:34 UTC
WHAT CHANGEDGitHub published CVE-2026-61439 on 7 October 2026, a HIGH-severity prompt injection vulnerability in PraisonAI that bypasses existing defenses. The first patched version is 4.6.78.
WHY IT MATTERSIf you run PraisonAI below 4.6.78, an attacker can inject prompts that circumvent the framework's built-in protections, potentially manipulating AI agent behavior or extracting sensitive context.
WHAT TO DOCheck your requirements.txt or pip freeze output for PraisonAI. If present and below 4.6.78, update to 4.6.78 or later in this patch cycle.
Model-reported confidence 85% (not measured accuracy)
Primary source ↗
written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+ghsa_id: GHSA-fj8f-m44g-c479 +summary: PraisonAI: Prompt Injection Defense Bypassed for HIGH-Level Threats +severity: high +published_at: 2026-10-07T20:43:58Z +cve_id: CVE-2026-61439 +vulnerabilities[].package.name: PraisonAI +vulnerabilities[].first_patched_version: 4.6.78
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
| triage verdict | MATERIAL at 0.78, needed 0.5 |
|---|---|
| triage reasoning | PraisonAI is an AI agent framework; prompt injection bypasses affect SDK security posture and customer data handling. |
| signals | AI framework package · prompt injection vulnerability · HIGH severity · first patched version specified · new advisory entry |
| suppression | — |
| triage model | anthropic/claude-haiku-4.5 ($0.00000) |
| writer model | anthropic/claude-sonnet-4.5 ($0.00000) |