// Security Advisory Triage

PraisonAI vulnerable to remote code execution via broken AST sandbox

GitHub advisories — pip · published 2026-10-31 05:47 UTC
ACTION NEEDED PATCH_NOW SECURITY
WHAT CHANGEDGitHub published GHSA-26mh-57q7-jfvr on 29 October 2026, rating a remote code execution vulnerability in PraisonAI as high severity. The vulnerability is tracked as CVE-2026-61438. Version 4.6.78 patches the issue.
WHY IT MATTERSAn attacker can execute arbitrary code remotely by exploiting the broken AST sandbox. Any system running an unpatched version of PraisonAI is exposed to full compromise.
WHAT TO DOUpgrade praisonai to version 4.6.78 or later. If you cannot upgrade immediately, verify whether you are running PraisonAI in production and isolate those instances from untrusted input.
Model-reported confidence 95% (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-26mh-57q7-jfvr
+summary: PraisonAI vulnerable to Remote Code Execution via Broken AST Sandbox
+severity: high
+published_at: 2026-10-29T16:34:26Z
+cve_id: CVE-2026-61438
+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 verdictMATERIAL at 0.92, needed 0.5
triage reasoningRCE in AI agent framework affecting dependency trees of services running AI workloads.
signalsknown exploitation: Remote Code Execution · AI framework: PraisonAI is an agent/SDK framework · high severity rating · patched version available: 4.6.78 · new advisory: first publication
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)