github.com PraisonAI Framework Vulnerability Allows Authentication Bypass
Article Content
- •CVE-2026-44338 allows unauthenticated access to PraisonAI workflows.
- •Attackers probed the vulnerability within hours of its disclosure.
- •Defenders should upgrade to version 4.6.34 and implement security best practices.
On May 11, 2026, a vulnerability (CVE-2026-44338) was disclosed in PraisonAI's multi-agent orchestration framework, where authentication was hard-coded to be disabled in versions 2.5.6 to 4.6.33. This flaw, with a CVSS score of 7.3, allows unauthorized access to agent workflows via the legacy Flask API. Attackers were able to probe a vulnerable instance just 3 hours and 44 minutes after the advisory was published, confirming the bypass with a 200 OK response on the /agents endpoint without an Authorization header. The vulnerability is compounded by other weaknesses, including CWE-668 and CWE-1188, and poses risks to internet-facing AI services. Organizations are advised to upgrade to version 4.6.34, bind servers to localhost, enable API keys, and monitor specific endpoints for reconnaissance activity. The incident highlights a broader trend of security oversights in AI service deployments.
Ask AI about this cluster
Answers cite the sources they use
Timeline
More articles in this cluster (4)
Following this threat?
Track Carbonato, Black Duck and CVE-2026-44338 in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.
Free account · no card needed