CVE-2026-72111

A flaw was found in the Linux kernel's BPF verifier. When processing a context load of an LSM hook return value, the markregs32range function incorrectly intersects new register bounds with stale ones, rather than replacing them. This can lead to a mismatch between the verifier's understanding of...

CVE-2026-72111

In the Linux kernel, the following vulnerability has been resolved: bpf: Reset register bounds before narrowing retval range in checkmemaccess When the BPF verifier processes a context load of an LSM hook return value, it calls markregs32range to narrow the register to the hook's valid range...

UBUNTU-CVE-2026-72111

In the Linux kernel, the following vulnerability has been resolved: bpf: Reset register bounds before narrowing retval range in checkmemaccess When the BPF verifier processes a context load of an LSM hook return value, it calls markregs32range to narrow the register to the hook's valid range...

CVE-2026-72111

CVE-2026-72111 affects the Linux kernel BPF verifier. The issue occurs when processing a context load of an LSM hook return value: __mark_reg_s32_range() narrows the destination register by intersecting with stale bounds instead of replacing them, potentially creating a verifier/runtime mismatch ...

Linux Distros Unpatched Vulnerability : CVE-2026-72111

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - bpf: Reset register bounds before narrowing retval range in checkmemaccess When the BPF verifier processes a context load of an LSM hook return value, it calls...

EUVD-2026-51922

A vulnerability in keras-team/keras versions = 3.14.0 allows arbitrary local HDF5 file content disclosure due to improper handling of HDF5 ExternalLinks. The KerasFileEditor and keras.saving.loadweights functions bypass the safegeth5group and safegeth5dataset helpers, which are designed to reject...

CVE-2026-9335 Improper Handling of HDF5 ExternalLinks in keras-team/keras

A vulnerability in keras-team/keras versions = 3.14.0 allows arbitrary local HDF5 file content disclosure due to improper handling of HDF5 ExternalLinks. The KerasFileEditor and keras.saving.loadweights functions bypass the safegeth5group and safegeth5dataset helpers, which are designed to reject...

CVE-2026-14534 Fickling check_safety() bypass via unlisted standard library modules (_posixsubprocess, site, atexit)

Trail of Bits fickling versions up to and including 0.1.10 do not include the Python standard library modules posixsubprocess, site, and atexit in the UNSAFEIMPORTS denylist fickle.py. Because these modules are absent from the denylist, fickling's checksafety function returns LIKELYSAFE with zero...

CVE-2025-71369 picklescan - Unsafe Deserialization via torch.utils.data.datapipes.utils.decoder.basichandlers

picklescan before 0.0.28 fails to detect malicious pickle files that use torch.utils.data.datapipes.utils.decoder.basichandlers in reduce methods, allowing attackers to bypass safety checks. Remote attackers can embed undetected malicious code in pickle files that executes during deserialization,...

CVE-2025-71369

picklescan before 0.0.28 fails to detect malicious pickle files that use torch.utils.data.datapipes.utils.decoder.basichandlers in reduce methods, allowing attackers to bypass safety checks. Remote attackers can embed undetected malicious code in pickle files that executes during deserialization,...

PT-2026-55676

Name of the Vulnerable Software and Affected Versions picklescan versions prior to 0.0.28 Description The software fails to detect malicious function calls to torch.utils.bottleneck. main .run cprofile within pickle files. This flaw allows remote attackers to bypass safety checks by embedding...

GuardFall Exposes Open-Source AI Coding Agents to Decades-Old Shell Injection Risks

The safety check that is supposed to stop an AI coding agent from running a dangerous command can be walked straight past using a shell trick that has been public for decades. New research from Adversa AI, which is named the bypass GuardFall , found it works against ten of the eleven popular...

around_render HTML-Safety Bypass

Summary ViewComponent::Basearoundrender can return HTML-unsafe strings that bypass the escaping behavior applied to normal call return values. This creates an XSS risk when downstream applications use aroundrender to wrap, replace, instrument, or conditionally return content that includes...

OS Command Injection

OliveTin is vulnerable to Command Injection. The vulnerability is due to insufficient input validation in Shell mode, where password-typed arguments and webhook-extracted JSON values bypass checkShellArgumentSafety before being passed to sh -c, allowing authenticated or unauthenticated attackers ...

CVE-2026-31700

A flaw was found in the Linux kernel. A Time-of-check to Time-of-use TOCTOU race condition exists in the tpacketsnd function when PACKETVNETHDR is enabled. A local user can exploit this by modifying the vnethdr fields in the mmap'd TX ring buffer between validation and use, thereby bypassing safe...

Attention Is Where You Attack

Safety-aligned large language models rely on RLHF and instruction tuning to refuse harmful requests, yet the internal mechanisms implementing safety behavior remain poorly understood. We introduce the Attention Redistribution Attack ARA, a white-box adversarial attack that identifies...

CVE-2026-41060

Summary: CVE-2026-41060 affects WWBN AVideo (versions 29.0 and below). The function isSSRFSafeURL() in objects/functions.php contains a same-domain shortcircuit (lines 4290-4296) that compares only hostname to webSiteRootURL and ignores the port, allowing an attacker to reach arbitrary ports on t...

CVE-2026-40149

PraisonAI is a multi-agent teams system. Prior to 4.5.128, the gateway's /api/approval/allow-list endpoint permits unauthenticated modification of the tool approval allowlist when no authtoken is configured the default. By adding dangerous tool names e.g., shellexec, filewrite to the allowlist, a...

EUVD-2026-21168

PraisonAI: Unauthenticated Allow-List Manipulation Bypasses Agent Tool Approval Safety Controls...

GHSA-4WR3-F4P3-5WJH PraisonAI: Unauthenticated Allow-List Manipulation Bypasses Agent Tool Approval Safety Controls

Summary The gateway's /api/approval/allow-list endpoint permits unauthenticated modification of the tool approval allowlist when no authtoken is configured the default. By adding dangerous tool names e.g., shellexec, filewrite to the allowlist, an attacker can cause the ExecApprovalManager to...