/

CVE-2026-57131

Severity:

CRITICAL RISK

(9.8)

Missing Authentication (CWE-306) in PraisonAI Jobs API

CVE-2026-57131

Deferred

Vulnerability type

Category:

Missing Authentication and Improper Authorization

CWE:

CWE-94, CWE-306, CWE-862

Impact:

Unauthorized Agent Execution and Sensitive Data Exposure

Severity:

CRITICAL RISK

(9.8)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

PraisonAI exposes its jobs API without authentication or per-job authorization, allowing network clients to submit attacker-controlled prompts and agent configurations. An attacker may execute unauthorized agents, access job results and service credentials, stream output, or cancel and delete other users' jobs.

Affected context

The vulnerability occurs in PraisonAI deployments where praisonai.jobs.server.create_app mounts the jobs router under /api/v1/runs without requiring identity verification or checking whether a caller is permitted to access a specific job. The risk is greatest when the service is reachable by untrusted or broadly trusted network clients and connected agents have access to credentials, tools, or sensitive data.

Exploit preconditions

  • Attack surface: The network-reachable /api/v1/runs jobs API. - Required access: Network connectivity to the service; no valid credentials

are required when the endpoint is exposed in the affected configuration.

  • Trigger point: Requests that create, enumerate, read, stream, cancel, or

delete jobs.

  • Why exploitation is possible: The API accepts attacker-controlled prompts

and agent settings while lacking both authentication and per-job ownership

checks, allowing callers to invoke agent capabilities and operate on jobs

belonging to other users.

Why this happens

The underlying flaw is an access-control failure in which a sensitive API is mounted without an authentication boundary and does not enforce object-level authorization. This maps primarily to CWE-306, Missing Authentication for Critical Function, with related behavior covered by CWE-862, Missing Authorization, and CWE-94, Improper Control of Generation of Code. Treating network reachability as sufficient trust allows unverified clients to supply agent instructions and control execution-related resources.

Potential impact

  • Unauthorized users may execute agents with attacker-controlled prompts or configurations.
  • Attackers could expose service credentials, job output, or data available to connected tools.
  • Unauthorized clients may stream, cancel, or delete jobs owned by other users.
  • Connected tools or downstream systems could be misused through agent execution.
  • The service could suffer data loss, resource exhaustion, or broader compromise depending on its privileges.

General mitigation guidance

Mitigation generally centers on placing a strong authentication boundary in front of every jobs endpoint and enforcing authorization for each job operation. Secure-by-default designs should require authenticated identities, verify job ownership or delegated permissions, and deny access when identity or policy information is absent. Reduce input trust by validating agent configuration against an allowlisted schema, limiting available tools and capabilities, and separating user prompts from privileged execution settings. Run agents with least privilege, isolate tenants and workloads, protect service credentials from agent contexts, and apply network segmentation and rate limits where appropriate.

How teams detect issues like this

Detection can include static analysis and code review checks for routers mounted without authentication dependencies, request handlers that omit identity checks, and job operations that use client-supplied identifiers without ownership validation. API policy testing should verify that anonymous callers cannot create, enumerate, read, stream, cancel, or delete jobs and that one authenticated tenant cannot access another tenant's resources. Runtime monitoring can flag anonymous requests, unusual job creation rates, unexpected tool invocation, credential access, and cross-user resource activity. Automated security review tools may help identify missing authentication and authorization enforcement.

Sources

SourceLink
github.comgithub.com/MervinPraison/PraisonAI/commit/2f9677abb2ea68eab864ee8b6a828fd0141612e1
github.comgithub.com/MervinPraison/PraisonAI/releases/tag/v4.6.58
github.comgithub.com/MervinPraison/PraisonAI/security/advisories/GHSA-fq2m-6wqh-x44g

Frequently asked

questions

Yes, when the jobs API is reachable by an attacker and remains exposed without an authentication and authorization boundary. Network access is generally sufficient; a separate account may not be required.

The attacker needs connectivity to the affected /api/v1/runs endpoints. The exposed deployment, network controls, and any external gateway determine whether that connectivity is available.

Require authentication for every jobs operation and enforce per-job authorization based on the authenticated identity. Also restrict agent tools, configuration, credentials, and runtime privileges to the minimum necessary.

Define authentication and authorization requirements at the API design stage, test anonymous and cross-tenant access paths, validate agent configurations against allowlists, and monitor privileged tool use and job activity.

Find out what's already
exploitable in your codebase.