Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems

As multi-agent systems enter high-stakes domains, the possibility that agents may circumvent safety boundaries is a growing concern. Prior work has examined this risk primarily in adversarial settings, where agents are instructed or rewarded to communicate covertly and evade oversight. We show th...

Memetic Trojans: Social Contagions As Carriers of Adversarial Payloads in Agent Networks

Autonomous large language model LLM agents increasingly interact in network environments where adversarial content can propagate between agents. Known attacks include agent worms, which spread through self-replicating prompt injections or configuration compromises. We introduce memetic trojans, a...

Xalgorix Autonomous AI Pentesting Agent 4.6.124

Most scanners detect. Xalgorix proves. An autonomous LLM agent works a full pentest methodology, then an independent verifier re-exploits every finding before it's reported - so you get proof, not a pile of maybes to triage. Self-hosted, private, and bring-your-own-LLM. Built in Go + TypeScript...

CVE-2026-13224: Improper Limitation of a Pathname to a Restricted Directory

A path traversal vulnerability in the Fireware OS WebUI management agent allows an authenticated administrator to read or list arbitrary files on the local filesystem by sending a specially crafted management request...

CVE-2026-100263: Improper Neutralization of Input During Web Page Generation

In JetBrains YouTrack before 2026.2.18991 stored HTML injection via the User-Agent header was possible...

MADBench: Benchmarking the Security of Multi-Agent Debate

Multi-agent debate MAD can improve large language model LLM reasoning by allowing multiple agents to exchange and critique their answers to the same task. However, the interactions that enable agents to correct mistakes can also spread adversarial errors and steer the agents toward an incorrect...

Aletheia: Permission-Minimality Testing for Coding-Agent Rules

Repository instruction files guide coding agents, but also expose them to prompt injection. Malicious rules can request credential access or data transfer while the agent produces a correct patch. We present Aletheia, a framework for permission-minimality testing. Aletheia translates requested...

Critical: amazon-ssm-agent

Issue Overview: Handshakes which used Encrypted Client Hello could be de-anonymized by a passive network observer due to a disclosure of pre-shared key identities in the unencrypted client hello. CVE-2026-42505 go-git is an extensible git implementation library written in pure Go. Prior to 5.19.2...

PT-2026-103101

Name of the Vulnerable Software and Affected Versions Post Views Stats Counter versions prior to 1.1.8 Description Stored Cross-Site Scripting occurs due to insufficient input sanitization and output escaping. Unauthenticated attackers can inject arbitrary web scripts through the User-Agent heade...

CVE-2026-51866

In DB-GPT 0.7.5 and 0.8.0, a skill uploaded through the real /api/v1/skills/upload route can later be executed through the real /api/v1/chat/react-agent flow...

CVE-2026-51866

Versions 0.7.5 through 0.8.0 of DB-GPT are vulnerable to a code injection issue (CWE-94). An attacker can upload a malicious skill via the /api/v1/skills/upload endpoint, which can subsequently be executed during the /api/v1/chat/react-agent flow. The impact includes potential compromise of confi...

CVE-2026-51852

The agent-zero software versions 1.7, 1.8, 1.9, and 1.10 are vulnerable to a Directory Traversal vulnerability within the python/helpers/file_browser.py component. Specifically, the FileBrowser.save_file_b64 method fails to perform proper normalization or validation on user-controlled file paths....

CVE-2026-51852

agent-zero 1.7, 1.8, 1.9, and 1.10 is vulnerable to Directory Traversal in python/helpers/filebrowser.py:FileBrowser.savefileb64. The savefileb64 method accepts user-controlled file paths without normalization or validation, allowing path traversal attacks...

PT-2026-103602

Name of the Vulnerable Software and Affected Versions agent-zero versions 1.7 through 1.10 Description A directory traversal issue exists in the save file b64 function within the python/helpers/file browser.py file. The function accepts user-controlled file paths without proper normalization or...

CVE-2026-51852: Improper Limitation of a Pathname to a Restricted Directory

agent-zero 1.7, 1.8, 1.9, and 1.10 is vulnerable to Directory Traversal in python/helpers/filebrowser.py:FileBrowser.savefileb64. The savefileb64 method accepts user-controlled file paths without normalization or validation, allowing path traversal attacks...

ALSA-2026:73765 Important: dogtag-pki security update

IdM PKI is an enterprise software system designed to manage enterprise Public Key Infrastructure deployments. IdM PKI consists of the following components: Certificate Authority CA Key Recovery Authority KRA Online Certificate Status Protocol OCSP Manager Token Key Service TKS Token Processing...

CVE-2026-51866: Improper Control of Generation of Code

In DB-GPT 0.7.5 and 0.8.0, a skill uploaded through the real /api/v1/skills/upload route can later be executed through the real /api/v1/chat/react-agent flow...

Speculative Safety Honeypot: Toward Proactive Defense against Multi-Turn Agent Attacks

As Large Language Model LLM agents are increasingly deployed in complex environments, multi-turn interaction attacks have become a significant security challenge. Existing detection methods typically rely on historical context. However, this retrospective logic struggles to identify deep maliciou...

CVE-2026-103057: Missing Authentication for Critical Function

AiSOC versions 5.1.0 before 12.0.0 contain an authentication bypass vulnerability in the realtime service internal endpoints POST /internal/agent-event and POST /internal/push. Attackers can post arbitrary events with spoofed tenant identifiers to broadcast malicious content over WebSocket and...

CentOS Linux 7 [TuxCare] Security Update: openssh / openssh-askpass / openssh-cavs / openssh-clients / etc (CENTOS7:CLSA-2025:1749824947)

The CentOS Linux 7 host has packages installed that are affected by a vulnerability as referenced in the TuxCare CENTOS7:CLSA-2025:1749824947 advisory. - In sshd in OpenSSH before 10.0, the DisableForwarding directive does not adhere to the documentation stating that it disables X11 and agent...