PrivateGPT < 0.5.0 - Open Redirect

An open redirect vulnerability exists in imartinez/privategpt version 0.5.0 due to improper handling of the 'file' parameter. This vulnerability allows attackers to redirect users to a URL specified by user-controlled input without proper validation or sanitization. id: CVE-2024-5936 info: name:...

GPT Academic v1.3.9 - Open Redirect

An open redirect vulnerability exists in GPT Academic v1.3.9, where the file parameter in the /file= endpoint can be manipulated to redirect users to malicious websites. This could facilitate phishing attacks by tricking users into visiting attacker-controlled URLs. id: CVE-2024-10812 info: name:...

CVE-2026-6684

CVE-2026-6684 affects FatFs prior to R0.16 when GPT scanning is used with FF_LBA64 = 1. The issue stems from an unbounded loop count derived from the GPT header field GPTH_PtNum, leading to extremely long or effectively infinite mount-time scans (CWE-835: Loop with Unreachable Exit Condition). Af...

PYSEC-2026-325 DB-GPT is vulnerable to SQL Injection attacks from unauthenticated users

In eosphoros-ai/db-gpt version v0.6.0, the web API POST /api/v1/editor/sql/run allows execution of arbitrary SQL queries without any access control. This vulnerability can be exploited by attackers to perform Arbitrary File Write using DuckDB SQL, enabling them to write arbitrary files to the...

PYSEC-2026-321 DB-GPT vulnerable to Arbitrary File Upload with Path Traversal

In eosphoros-ai/db-gpt version v0.6.0, the web API POST /v1/personal/agent/upload is vulnerable to Arbitrary File Upload with Path Traversal. This vulnerability allows unauthorized attackers to upload arbitrary files to the victim's file system at any location. The impact of this vulnerability...

PYSEC-2026-322 DB-GPT Arbitrary File Write vulnerability

In eosphoros-ai/db-gpt version v0.6.3 and earlier, the web API POST /api/v1/editor/chart/run allows execution of arbitrary SQL queries without any access control. This vulnerability can be exploited by attackers to perform Arbitrary File Write, enabling them to write arbitrary files to the victim...

PYSEC-2026-323 DB-GPT Absolute Path Traversal vulnerability

In eosphoros-ai/db-gpt version 0.6.0, the endpoint for uploading files is vulnerable to absolute path traversal. This vulnerability allows an attacker to upload arbitrary files to arbitrary locations on the target server. The issue arises because the filekey and docfile.filename parameters are...

OpenAI Previews GPT-5.6 Sol With Restricted Access and Stronger Cyber Safeguards

OpenAI on Friday released three versions of GPT-5.6 , called Sol, Terra, and Luna , as a limited preview to a small number of companies as part of an ongoing engagement with the U.S. government. While Sol is the latest flagship model and the most powerful, Terra strikes a balance between efficien...

CVE-2025-32437

CVE-2025-32437 affects AutoGPT prior to 0.6.63, specifically the MediaDurationBlock. The issue arises because MediaDurationBlock downloads and stores videos in a temporary directory without proper deletion, and StepThroughItemsBlock can iterate MediaDurationBlock multiple times, with no limit on ...

Malicious code in gpt-sdk (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector 8b9bdc5e04979d5b4f73407bcedaecc9df24dbb03e0bfbc0edefe333023dc50c On npm install, postinstall.js runs unconditionally and collects a wide range of installer-side reconnaissance data: hostname and FQDN, contents of...

CVE-2026-45567 Roxy-WI: Authentication bypass via 'api' substring in URL + unauthenticated /api/gpt

Roxy-WI is a web interface for managing Haproxy, Nginx, Apache and Keepalived servers. In versions 8.2.6.4 and prior, there is an authentication bypass vulnerability via 'api' substring in URL + unauthenticated /api/gpt. At time of publication, there are no publicly available patches...

CVE-2026-45567 Roxy-WI: Authentication bypass via 'api' substring in URL + unauthenticated /api/gpt

Roxy-WI is a web interface for managing Haproxy, Nginx, Apache and Keepalived servers. In versions 8.2.6.4 and prior, there is an authentication bypass vulnerability via 'api' substring in URL + unauthenticated /api/gpt. At time of publication, there are no publicly available patches...

CVE-2026-45567

Roxy-WI is a web interface for managing HAProxy, Nginx, Apache and Keepalived. In versions 8.2.6.4 and prior, there is an authentication bypass via the URL containing the substring 'api' and an unauthenticated /api/gpt path. The CVSS v3.1 base score is 8.3 (HIGH) with NETWORK attack vector and no...

Arm: Completion of memory accesses not guaranteed by completion of a TLBI

ISSUE DESCRIPTION A hardware issue has been identified in certain Arm CPU designs. A broadcast TLBI on one PE may complete before affected memory accesses on another PE are globally observed. This may permit bypass of Stage 1 translation, Stage 2 translation, or GPT protection. The erratum occurs...

CVE-2026-33233

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. In versions 0.6.34 through 0.6.51, the backend deserializes Redis cache bytes using pickle.loads without integrity/authenticity checks. The write path serializes values with...

Empirical Evaluation of Large Language Models for Migration of Code Fragments to Post-Quantum Cryptography

The transition to post-quantum cryptography PQC requires not only replacing vulnerable cryptographic primitives, but also refactoring the surrounding software logic. While existing PQC migration frameworks provide organizational guidance, practical code-level remediation remains largely manual an...

CVE-2026-10566 FoundationAgents MetaGPT schema.py Message.check_instruct_content deserialization

A weakness has been identified in FoundationAgents MetaGPT up to 0.8.2. This affects the function Message.checkinstructcontent of the file metagpt/schema.py. Executing a manipulation of the argument mapping can lead to deserialization. The attack is restricted to local execution. The exploit has...

EUVD-2026-33872

A weakness has been identified in FoundationAgents MetaGPT up to 0.8.2. This affects the function Message.checkinstructcontent of the file metagpt/schema.py. Executing a manipulation of the argument mapping can lead to deserialization. The attack is restricted to local execution. The exploit has...

A New Framework for Cybersecurity Refusals in AI Agents

Agentic scaffolds have dramatically improved LLM performance on complex, long-horizon tasks, yielding both broad benefits and amplified risks in domains like cybersecurity. Existing benchmarks for AI agents in cybersecurity focus mainly on measuring proficiency--how effectively agents can complet...

Quality-Diversity Evolution for Discovering Diverse Vulnerabilities in LLM Safety

Current approaches to LLM adversarial testing suffer from coverage gaps: manual red-teaming does not scale, LLM-as-attacker methods exhibit mode collapse, and gradient-based approaches produce uninterpretable gibberish. We introduce a quality-diversity evolutionary framework that operates at the...