Astra Linux – Vulnerability in Firefox

An attacker could write data to the user’s clipboard by bypassing the user prompt during a certain sequence of navigation events. This vulnerability affects Firefox 129, Firefox ESR 128.3, and Thunderbird 128.3...

Astra Linux – Vulnerability in Firefox

A ambiguous file picker design could have confused users who intended to select and upload a single file, into uploading an entire directory. This issue was addressed by adding a new prompt. This vulnerability affects Firefox versions earlier than 85...

Astra Linux – Vulnerability in Firefox

When a user has already allowed a website to access the microphone and camera, disabling camera sharing does not completely prevent the website from re-enabling them without an additional prompt. This is only possible if the website continues to record with the microphone until the camera is...

Astra Linux – Vulnerability in Thunderbird

The ms-msdt, search, and search-ms protocols deliver content to Microsoft applications, bypassing the browser, when a user accepts a prompt. These applications have known vulnerabilities that have been exploited in the wild although we know of none exploited through Thunderbird. Therefore, in thi...

Astra Linux – Vulnerability in zsh

In zsh before version 5.8.1, an attacker can gain code execution if they control the command output within the prompt, as demonstrated by using a %F argument. This occurs due to the recursive PROMPTSUBST expansion...

Astra Linux – Vulnerability in Firefox and Thunderbird

Documents in deeply-nested cross-origin browsing contexts may have obtained permissions granted to the top-level origin, bypassing the existing prompts and wrongly inheriting the top-level permissions. This vulnerability affects Thunderbird 91.9, Firefox ESR 91.9, and Firefox 100...

Astra Linux – Vulnerability in Firefox

Even when an iframe was sandboxed with allow-top-navigation-by-user-activation, if it received a redirect header to an external protocol, the browser would process the redirect and prompt the user appropriately. This vulnerability affects Firefox versions earlier than 102...

Astra Linux – Vulnerability in Firefox and Thunderbird

Through a series of popups that reuse windowName, an attacker can cause a window to go fullscreen without the user seeing the notification prompt. This results in potential user confusion or spoofing attacks. This vulnerability affects Firefox ESR 102.5, Thunderbird 102.5, and Firefox 107...

Astra Linux – Vulnerability in Firefox and Thunderbird

By displaying a form validation message in the correct location at the same time as a permission prompt such as for geolocation, the validation message could potentially obscure the prompt, allowing the user to be tricked into granting the permission. This vulnerability affects Firefox 94,...

Astra Linux – Vulnerability in Chromium

Inappropriate implementation in Prompts in Google Chrome prior to 117.0.5938.62 allowed a remote attacker to spoof security UI via a crafted HTML page. Chromium security severity: Medium...

Astra Linux – Vulnerability in Firefox

A malicious website can enter fullscreen mode while simultaneously triggering a WebAuthn prompt. This could have obscured the fullscreen notification and could have been exploited in a spoofing attack. This vulnerability affects Firefox versions earlier than 119...

Astra Linux – Vulnerability in Firefox and Thunderbird

The permission prompt input delay may expire if the window is not in focus. This makes it vulnerable to clickjacking by malicious websites. This vulnerability affects Firefox 124, Firefox ESR 115.10, and Thunderbird 115.10...

CVE-2026-57129: Improper Limitation of a Pathname to a Restricted Directory

PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, MentionsParser.processfilemention accepts file-mention values and falls back from workspace-relative resolution to Pathfilepath without traversal, symlink, or workspace-boundary validation. Prompt input from users, bots, or...

PT-2026-91650

Name of the Vulnerable Software and Affected Versions macOS versions prior to Golden Gate 27 Description A permissions issue exists due to improper state management, which could allow an application to bypass Apple Intelligence security prompts. Recommendations Update to macOS Golden Gate 27...

PIDS-Bench: Evaluating Prompt-Injection Detectors under Over-Defense, Obfuscation, and Distribution Shift

Prompt-injection detectors are typically evaluated using aggregate F1 on in-distribution test data, which offers limited insight into behavior under distribution shift, particularly on the benign side of the decision boundary, where false positives impose direct operational cost yet are seldom...

SkillSecurer: Detecting and Patching Prompt-Injection Vulnerabilities in AI Agent Skills

Agent skills extend AI agents with reusable instructions, scripts, and configuration, but are also open to new attacks to influence an agent's decisions and actions. To address these risks, we present SkillSecurer, a fully agentic framework for generating, detecting, localising, and remediating...

sage v0.13.0

Sage Safety for Agents — Agent Detection & Response for AI coding assistants Sage is a lightweight security layer that protects AI agents from executing dangerous actions. It intercepts tool calls — shell commands, URL fetches, file writes — and checks them against multiple threat detection layer...

Anthropic Says Seven China-Based AI Labs Ran Industrial-Scale Claude Distillation Attacks

Anthropic on Thursday said it identified and disrupted industrial-scale illicit distillation attacks against Claude from seven labs based in China, including Alibaba, Moonshot, DeepSeek, Z.ai aka Zhipu, and MiniMax. Knowledge distillation by itself is a legitimate training method. It refers to a...

garak v0.17.0

garak, LLM vulnerability scanner Generative AI Red-teaming & Assessment Kit garak checks if an LLM can be made to fail in a way we don't want. garak probes for hallucination, data leakage, prompt injection, misinformation, toxicity generation, jailbreaks, and many other weaknesses. If you know nm...

PYSEC-2026-3938 vLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts

Executive Summary The follow-up protection for CVE-2025-62164 is incomplete at vLLM revision 26587f9519e22a5c4549ead7595ad9ca3229c4fd. It wraps serialized prompt-embedding reconstruction and dense conversion in torch.sparse.checksparsetensorinvariants, but PyTorch 2.11.0 implements that context...