CVE-2026-33234 AutoGPT: SendEmailBlock's IP blocklist bypass allows SSRF via user-controlled SMTP server

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. In versions 0.1.0 through 0.6.51, SendEmailBlock in autogptplatform/backend/backend/blocks/emailblock.py accepts a user-supplied smtpserver string and smtpport integer as...

CVE-2026-33233

CVE-2026-33233 affects AutoGPT Platform: older releases (0.6.34–0.6.51) deserialize Redis cache bytes with pickle.loads without integrity checks, while writes use pickle.dumps into Redis. The read path blindly calls pickle.loads on bytes with no HMAC/signature or strict schema validation. An atta...

CVE-2026-33232

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Versions 0.4.2 through 0.6.51 are vulnerable to an unauthenticated Denial of Service DoS through the server due to uncontrolled disk space consumption. The downloadagentfile...

EUVD-2026-30819

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Versions 0.4.2 through 0.6.51 are vulnerable to an unauthenticated Denial of Service DoS through the server due to uncontrolled disk space consumption. The downloadagentfile...

MAL-2026-4020 Malicious code in @antv/gpt-vis (npm)

Part of the Mini Shai-Hulud supply chain attack campaign in which a threat actor compromised the npm account atool and published 631 malicious versions across 314 npm packages in an automated 22-minute burst. Each malicious version injects a preinstall hook that executes a 498KB obfuscated Bun...

Malicious code in @antv/gpt-vis-ssr (npm)

Part of the Mini Shai-Hulud supply chain attack campaign in which a threat actor compromised the npm account atool and published 631 malicious versions across 314 npm packages in an automated 22-minute burst. Each malicious version injects a preinstall hook that executes a 498KB obfuscated Bun...

MAL-2026-4021 Malicious code in @antv/gpt-vis-ssr (npm)

Part of the Mini Shai-Hulud supply chain attack campaign in which a threat actor compromised the npm account atool and published 631 malicious versions across 314 npm packages in an automated 22-minute burst. Each malicious version injects a preinstall hook that executes a 498KB obfuscated Bun...

CVE-2025-32425

AutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows. In AutoGPT, the execution process is recorded to the console stdout/stderr, and deployed in container mode, which is automatically captured by Docker an...

CVE-2025-32425

CVE-2025-32425 affects AutoGPT platform prior to v0.6.32, where container execution logs emitted to stdout/stderr could be captured by Docker and stored as container logs without a size limit. This lack of log rotation/log size control can lead to server disk resource exhaustion and DoS under hig...

OpenAI’s GPT-5.5 is as Good as Mythos at Finding Security Vulnerabilities

The UK's AI Security Institute evaluated GPT-5.5's ability to find security vulnerabilities, and found that it is comparable to Claude Mythos. Note that the OpenAI model is generally available. Here is the Institute's evaluation of Mythos. And here is an analysis of a smaller, cheaper model. It...

CVE-2026-31246

GPT-Pilot thru commit 0819827ce20346ef5f25b3fe29293cb448840565 2025-09-03 contains a command injection vulnerability CWE-78 in the Executor.run method. During project execution, when the system prompts the user to confirm or modify a command to be run, it accepts free-text input without proper...

Command Injection

Overview gpt-pilot is a GPT Pilot - an AI developer that works with you to build complex projects Affected versions of this package are vulnerable to Command Injection via the Executor.run function. An attacker can execute arbitrary shell commands by supplying crafted input that is passed directl...

GHSA-M85W-WHWH-QVFX GPT-Pilot contains a command injection vulnerability in the Executor.run() method

GPT-Pilot thru commit 0819827ce20346ef5f25b3fe29293cb448840565 2025-09-03 contains a command injection vulnerability CWE-78 in the Executor.run method. During project execution, when the system prompts the user to confirm or modify a command to be run, it accepts free-text input without proper...

CVE-2026-31246

GPT-Pilot thru commit 0819827ce20346ef5f25b3fe29293cb448840565 2025-09-03 contains a command injection vulnerability CWE-78 in the Executor.run method. During project execution, when the system prompts the user to confirm or modify a command to be run, it accepts free-text input without proper...

CVE-2026-31246

GPT-Pilot thru commit 0819827ce20346ef5f25b3fe29293cb448840565 2025-09-03 contains a command injection vulnerability CWE-78 in the Executor.run method. During project execution, when the system prompts the user to confirm or modify a command to be run, it accepts free-text input without proper...

CVE-2026-31246

GPT-Pilot thru commit 0819827ce20346ef5f25b3fe29293cb448840565 2025-09-03 contains a command injection vulnerability CWE-78 in the Executor.run method. During project execution, when the system prompts the user to confirm or modify a command to be run, it accepts free-text input without proper...

PT-2026-39616

Name of the Vulnerable Software and Affected Versions GPT-Pilot versions prior to commit 0819827ce20346ef5f25b3fe29293cb448840565 Description Command injection occurs in the Executor.run method. During project execution, the system accepts free-text input for confirming or modifying commands...

CVE-2026-31246

GPT-Pilot (through commit 0819827ce20346ef5f25b3fe29293cb448840565) contains a command injection vulnerability (CWE-78) in the Executor.run() path. User prompts to confirm/modify a command are accepted as free-text and directly passed to asyncio.create_subprocess_shell(), enabling an attacker to ...

EUVD-2026-26258

DocsGPT is a GPT-powered chat for documentation. From version 0.15.0 to before version 0.16.0, an attacker accessing both the official DocsGPT website or any local and public deployment, can craft a malicious payload bypassing the "MCP test" behavior to achieve arbitrary remote code execution RCE...

Jailbreaking Frontier Foundation Models through Intention Deception

Large vision-language models exhibit remarkable capability but remain highly susceptible to jailbreaking. Existing safety training approaches aim to have the model learn a refusal boundary between safe and unsafe, based on the user's intent. It has been found that this binary training regime ofte...