Critical CVEs
New CVEs as they are published, with CVSS score, CISA KEV status, public exploits and the vendor's fix. Worth tracking, but only a small part of what Panop does: mapping what you expose to the internet, testing what an attacker can actually exploit, ranking it by business impact, and helping your team understand and fix it.
Feed at a glance
- Tracked
- 13360
- Critical
- 2486
- CISA KEV
- 125
- Public PoC
- 437
Most common types
- Broken Access Control 1691
- Remote Code Execution 1609
- Privilege Escalation 1307
- Authentication Bypass 1121
- Denial of Service 799
Published per day, last 14 days
Source: public CVE records (NVD, CISA KEV) · last published 3 Nov 2026
Critical CVEs
Newest first · severity as published by the vendor or NVD · page 144 of 166
- 9.8 /10 CriticalCVE-2026-63695
Dell SmartFabric OS10 Software - Authentication Bypass
Dell SmartFabric OS10 Software < 10.6.1.3 contains a session fixation vulnerability caused by improper session management, letting unauthenticated remote attackers steal sessions, exploit requires remote access.
- 9.8 /10 CriticalCVE-2026-62379 No auth
Open Access Management (OpenAM) - Remote Code Execution
Open Access Management (OpenAM) < 16.1.2 contains a remote code execution caused by unsafe deserialization and arbitrary class instantiation in the pre-authentication /authservice PLL endpoint, letting unauthenticated attackers execute code remotely.
- 9.9 /10 CriticalCVE-2026-61667 Auth required
DIRAC - Command Injection
DIRAC < 8.0.79, 9.0.22, 9.1.10 contains a command injection caused by unparameterized SQL query in DatasetManager.py __checkDataset, letting authenticated attackers execute arbitrary commands, exploit requires authentication.
- 9.6 /10 CriticalCVE-2026-61568 No auth
@zereight/mcp-gitlab - DNS Rebinding
@zereight/mcp-gitlab < 2.1.30 contains a dns rebinding vulnerability caused by lack of effective Host or Origin allowlist in the Streamable HTTP MCP endpoint, letting attackers route browser requests to local MCP listener with attacker-controlled headers, exploit requires victim to visit malicious web page.
- 9.8 /10 CriticalCVE-2026-61560 No auth
@zereight/mcp-gitlab - Information Disclosure
@zereight/mcp-gitlab < 2.1.27 contains an information disclosure caused by unsanitized file_path parameter in upload_markdown tool with SSE=true, letting unauthenticated network attackers read arbitrary files and steal GitLab tokens, exploit requires default SSE transport mode enabled.
- 9.6 /10 CriticalCVE-2026-61559 Auth required
@zereight/mcp-gitlab - Information Disclosure
@zereight/mcp-gitlab < 2.1.27 contains an information disclosure caused by lack of hostname restriction on X-GitLab-API-URL header, letting attackers receiving victim's Private-Token via attacker-controlled API URL, exploit requires network access to send crafted requests.
- 10.0 /10 CriticalCVE-2026-59971 No auth
MySQL MCP Server - DNS Rebinding
MySQL MCP Server < 0.4.2 contains a DNS rebinding and unauthenticated access vulnerability caused by insecure SseServerTransport configuration and lack of authentication on key routes, letting network attackers disclose and modify database contents and potentially execute code if FILE privileges exist, exploit requires network access and victim browser interaction.
- 9.8 /10 CriticalCVE-2026-57148 No auth
PraisonAI praisonai-platform - Authentication Bypass
PraisonAI praisonai-platform < 0.1.6 contains a broken authentication caused by fallback to a public HS256 signing key and disabled guards, letting unauthenticated attackers impersonate users or workspace owners, exploit requires knowledge of target identifier.
- 9.8 /10 CriticalCVE-2026-57147 No auth
PraisonAI praisonai-platform - Authentication Bypass
PraisonAI praisonai-platform < 0.1.6 contains a broken authentication caused by default JWT_SECRET and missing production guard in auth_service.py, letting remote unauthenticated attackers forge tokens and bypass authentication, exploit requires PLATFORM_JWT_SECRET and PLATFORM_ENV unset.
- 9.8 /10 CriticalCVE-2026-57141 No auth
PraisonAI - Command Injection
PraisonAI < 1.7.2 contains a command injection caused by execution of model-generated JavaScript using new Function() and with(sandbox), letting attackers execute OS commands and access host process capabilities, exploit requires attacker to influence the code argument.
- 9.4 /10 CriticalCVE-2026-57140 No auth
PraisonAI - Authentication Bypass
PraisonAI 1.6.0 to 1.7.2 contains an authentication bypass caused by missing authentication middleware on GET /api/agents and POST /api/chat endpoints, letting remote attackers access agent data and invoke agents, exploit requires network access to the service.
- 9.8 /10 CriticalCVE-2026-57139
PraisonAI - Authentication Bypass
PraisonAI 1.5.0 to 1.7.2 contains an authentication bypass caused by unrestricted HTTP POST requests in MCPServer.startHttp(), letting remote attackers execute server-side handlers with elevated privileges or disclose data, exploit requires network access to the port.
- 9.9 /10 CriticalCVE-2026-57138 Auth required
PraisonAI - Remote Code Execution
PraisonAI 1.4.0 to 1.7.2 contains a remote code execution caused by insecure sandboxing in codeMode using new Function() and bypassing restrictions, letting attackers execute arbitrary code and access host resources, exploit requires control of codeMode input.
- 9.8 /10 CriticalCVE-2026-56960 No auth
Unknown - Use After Free
Unknown product contains a use-after-free caused by a logic error in multiple locations, letting remote attackers escalate privileges without additional execution privileges, exploit requires no user interaction.
- 9.8 /10 CriticalCVE-2026-55366 No auth
IP Multimedia Subsystem - Authentication Bypass
IP Multimedia Subsystem contains an authentication bypass caused by a logic error in the code, letting remote attackers escalate privileges without additional execution privileges, exploit requires no user interaction.