CVE-2026-47393

CVE-2026-44338 (GHSA-6rmh-7xcm-cpxj) documents that PraisonAI ships a code-generator (praisonai.deploy.api.generate_api_server_code) that emits a Flask API server with authentication disabled by default. Users who follow the documented quickstart (praisonai deploy --type api) get a server that:

• binds to 0.0.0.0 per the recommended sample YAML
• exposes /chat and /agents endpoints
• runs praisonai.run() on user-supplied JSON input, LLM orchestration with the API key materials present in the process environment
• does not require any authentication

The PyPI wheel praisonai==4.6.33 (current @latest) still ships the generator with auth_enabled defaulting to False. The fix shape is opt-in via APIConfig(auth_enabled=True, auth_token=...).

Details

Anchor (file:line:symbol)

• Vulnerable artifact: praisonai==4.6.33 on PyPI.
• Defaults: praisonai/deploy/models.py:29, auth_enabled: bool = Field(default=False, ...); praisonai/deploy/models.py:30, auth_token: Optional[str] = Field(default=None, ...).
• Generator: praisonai/deploy/api.py:40, AUTH_ENABLED = {config.auth_enabled}; api.py:41, AUTH_TOKEN = {repr(config.auth_token)}; api.py:43-49, def check_auth(): if not AUTH_ENABLED: return True.
• CLI entry: documented as praisonai deploy --type api (vendor README); produces the generator output above with no flag required to suppress the warning, because no warning is emitted.

Vulnerable code (verbatim from installed wheel)

# praisonai/deploy/models.py (praisonai==4.6.33)
class APIConfig(BaseModel):
    host: str = Field(default="127.0.0.1", description="Server host")
    port: int = Field(default=8005, description="Server port")
    cors_enabled: bool = Field(default=True, description="Enable CORS")
    auth_enabled: bool = Field(default=False, description="Enable authentication")     # line 29
    auth_token: Optional[str] = Field(default=None, description="Authentication token") # line 30

# praisonai/deploy/api.py (praisonai==4.6.33)
code = f\'\'\'...
# Authentication
AUTH_ENABLED = {config.auth_enabled}      # False by default
AUTH_TOKEN   = {repr(config.auth_token)}  # None by default

def check_auth():
    if not AUTH_ENABLED:
        return True                       # short-circuit, accept all
    token = request.headers.get(\'Authorization\', \'\').replace(\'Bearer \', \'\')
    return token == AUTH_TOKEN
...
\'\'\'

A default invocation of the deploy command emits a server whose check_auth() short-circuits to True and accepts unauthenticated /chat, /agents POSTs.

PoC

#!/usr/bin/env python3
"""
legend-c420 PoC - PraisonAI 4.6.33 generates Flask API server with auth
disabled by default. Class H sibling of CVE-2026-44338.

Phase 1: reflect on praisonai.deploy.models.APIConfig defaults.
Phase 2: call generate_api_server_code(default config) and assert the
         emitted source contains AUTH_ENABLED = False and the
         short-circuit return.
Phase 3: re-run with auth_enabled=True, auth_token='s3cret-bearer-value'
         and confirm the emitted source flips to the secure shape.

Exit code 0 = PASS = vulnerable defaults confirmed.
"""
import sys, traceback

def phase1_dataclass_defaults():
    print("PHASE 1 - praisonai.deploy.models.APIConfig default values")
    from praisonai.deploy.models import APIConfig
    cfg = APIConfig()
    checks = [
        ("auth_enabled", cfg.auth_enabled, False),
        ("auth_token",   cfg.auth_token,   None),
    ]
    for name, observed, expected in checks:
        ok = observed == expected
        mark = "VULNERABLE" if name in ("auth_enabled","auth_token") and ok else "ok"
        print(f"  {name:14s} = {observed!r:18s}  (expected {expected!r})  [{mark}]")
        assert ok
    print("  >> APIConfig defaults reproduce the CVE-2026-44338 shape.")

def phase2_default_generator_emits_unauth():
    print("PHASE 2 - generate_api_server_code(default config) emits unauth server")
    from praisonai.deploy.models import APIConfig
    from praisonai.deploy.api import generate_api_server_code
    src = generate_api_server_code("agents.yaml", config=APIConfig())
    for needle in ["AUTH_ENABLED = False","AUTH_TOKEN = None","if not AUTH_ENABLED:","return True"]:
        assert needle in src, f"missing: {needle!r}"
        print(f"  [FOUND] {needle!r}")
    print("  >> Default-config generator emits Flask server with check_auth() short-circuit.")

def phase3_fix_shape_available():
    print("PHASE 3 - auth_enabled=True flips to secure shape")
    from praisonai.deploy.models import APIConfig
    from praisonai.deploy.api import generate_api_server_code
    cfg = APIConfig(auth_enabled=True, auth_token="s3cret-bearer-value")
    src = generate_api_server_code("agents.yaml", config=cfg)
    assert "AUTH_ENABLED = True" in src
    assert "AUTH_ENABLED = False" not in src
    print("  >> Fix shape works when toggled. Class H confirmed: default is insecure.")

def main():
    print("=" * 64)
    print("legend-c420 PoC - PraisonAI default-config AUTH_ENABLED=False")
    print("=" * 64)
    try:
        phase1_dataclass_defaults()
        phase2_default_generator_emits_unauth()
        phase3_fix_shape_available()
    except Exception:
        traceback.print_exc()
        print("FAIL"); sys.exit(2)
    print("PASS 3/3 phases. EXIT 0.")
    sys.exit(0)

if __name__ == "__main__":
    main()

PoC dependencies: praisonai==4.6.33 from PyPI. Tested on Python 3.11.

Run log verdict: PASS 3/3 phases. EXIT 0., vulnerable-default shape confirmed. auth_enabled=False by default, check_auth() short-circuits to True, fix toggle exists but is opt-in.

An operator who runs the vendor-documented quickstart (pip install praisonai && praisonai deploy --type api) gets a network-reachable Flask server that invokes praisonai.run() on attacker-supplied JSON with the user's LLM API keys in the process environment. The attacker reaches arbitrary LLM-orchestration (including any tool-use the agents define, which in PraisonAI commonly includes python_repl, bash, file I/O, and HTTP calls), with the host's API-key credit billed to the operator.

• Belief: CVE-2026-44338 was filed and triaged.
• Reality: praisonai==4.6.33 is current @latest on PyPI (2026-05-16). The generator still defaults to auth_enabled=False.
• Gap: The CVE acknowledges the fix shape exists. The fix is opt-in. The default-config consumer remains vulnerable.

Parent CVE: CVE-2026-44338 / GHSA-6rmh-7xcm-cpxj

A critical operation is accessible without requiring any authentication. Typical impact: any user can invoke the privileged function.

CVE-2026-47393 has a CVSS score of 9.8 (Critical). The vector is network-reachable, no privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (4.6.40); upgrading removes the vulnerable code path.

Kodem intelligence

1. What is CVE-2026-47393? CVE-2026-47393 is a critical-severity missing authentication for critical function vulnerability in PraisonAI (pip), affecting versions <= 4.6.39. It is fixed in 4.6.40. A critical operation is accessible without requiring any authentication.
2. How severe is CVE-2026-47393? CVE-2026-47393 has a CVSS score of 9.8 (Critical). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
3. Which versions of PraisonAI are affected by CVE-2026-47393? PraisonAI (pip) versions <= 4.6.39 is affected.
4. Is there a fix for CVE-2026-47393? Yes. CVE-2026-47393 is fixed in 4.6.40. Upgrade to this version or later.
5. Is CVE-2026-47393 exploitable, and should I be worried? Whether CVE-2026-47393 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
6. What actually determines whether CVE-2026-47393 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
7. How do I fix CVE-2026-47393? Upgrade PraisonAI to 4.6.40 or later.