#git

# Summary vLLM loads Hugging Face `auto_map` dynamic modules during model resolution **without gating on `trust_remote_code`**, allowing attacker-controlled Python code in a model repo/path to execute at server startup. --- # Impact An attacker who can influence the model repo/path (local directory or remote Hugging Face repo) can achieve **arbitrary code execution** on the vLLM host during model load. This happens **before any request handling** and does **not require API access**. --- # Affected Versions All versions where `vllm/model_executor/models/registry.py` resolves `auto_map` entries with `try_get_class_from_dynamic_module` **without checking `trust_remote_code`** (at least current `main`). --- # Details During model resolution, vLLM unconditionally iterates `auto_map` entries from the model config and calls `try_get_class_from_dynamic_module`, which delegates to Transformers’ `get_class_from_dynamic_module` and **executes the module code**. This occurs even when ...

Zoom and GitLab have released security updates to resolve a number of security vulnerabilities that could result in denial-of-service (DoS) and remote code execution. The most severe of the lot is a critical security flaw impacting Zoom Node Multimedia Routers (MMRs) that could permit a meeting participant to conduct remote code execution attacks. The vulnerability, tracked as CVE-2026-22844

Due to improper input validation, a malicious object key can lead to prototype pollution during JSON deserialization. This affects only JSON deserialization functionality. As there is no known workaround, please upgrade to the latest version.

### Impact This vulnerability affects Laravel Reverb versions prior to v1.7.0 when horizontal scaling is enabled (`REVERB_SCALING_ENABLED=true`). The exploitability of this vulnerability is increased because Redis servers are commonly deployed without authentication. With horizontal scaling enabled, Reverb servers communicate via Redis PubSub. Reverb previously passed data from the Redis channel directly into PHP’s `unserialize()` function without restricting which classes could be instantiated. **Risk:** Remote Code Execution (RCE) ### Patches This vulnerability is fixed in Laravel Reverb v1.7.0. Update your dependency to `laravel/reverb: ^1.7.0` immediately. ### Workarounds If you cannot upgrade to v1.7.0, you should apply the following mitigations: * Redis Security: Require a strong password for Redis access and ensure the service is only accessible via a private network or local loopback. * Disable Scaling: If your environment uses only one Reverb node, set `REVERB_SCALING_...

Deployments of Apache Solr 5.3.0 through 9.10.0 that rely on Solr's "Rule Based Authorization Plugin" are vulnerable to allowing unauthorized access to certain Solr APIs, due to insufficiently strict input validation in those components.  Only deployments that meet all of the following criteria are impacted by this vulnerability: * Use of Solr's "RuleBasedAuthorizationPlugin" * A RuleBasedAuthorizationPlugin config (see security.json) that specifies multiple "roles" * A RuleBasedAuthorizationPlugin permission list (see security.json) that uses one or more of the following pre-defined permission rules: "config-read", "config-edit", "schema-read", "metrics-read", or "security-read". * A RuleBasedAuthorizationPlugin permission list that doesn't define the "all" pre-defined permission * A networking setup that allows clients to make unfiltered network requests to Solr. (i.e. user-submitted HTTP/HTTPS requests reach Solr as-is, unmodified or restricted by any intervening pro...

The "create core" API of Apache Solr 8.6 through 9.10.0 lacks sufficient input validation on some API parameters, which can cause Solr to check the existence of and attempt to read file-system paths that should be disallowed by Solr's "allowPaths" security setting https://https://solr.apache.org/guide/solr/latest/configuration-guide/configuring-solr-xml.html#the-solr-element .  These read-only accesses can allow users to create cores using unexpected configsets if any are accessible via the filesystem.  On Windows systems configured to allow UNC paths this can additionally cause disclosure of NTLM "user" hashes.  Solr deployments are subject to this vulnerability if they meet the following criteria: * Solr is running in its "standalone" mode. * Solr's "allowPath" setting is being used to restrict file access to certain directories. * Solr's "create core" API is exposed and accessible to untrusted users.  This can happen if Solr's RuleBasedAuthorizationPlugin https://solr.ap...

A flaw was found in the Keycloak Admin REST API. This vulnerability allows the exposure of backend schema and rules, potentially leading to targeted attacks or privilege escalation via improper access control.

Boston, MA, USA, 21st January 2026, CyberNewsWire

Researchers showed how prompt injection hidden in a calendar invite can bypass privacy controls and turn an AI assistant into a data-leaking accomplice.

Every managed security provider is chasing the same problem in 2026 — too many alerts, too few analysts, and clients demanding “CISO-level protection” at SMB budgets. The truth? Most MSSPs are running harder, not smarter. And it’s breaking their margins. That’s where the quiet revolution is happening: AI isn’t just writing reports or surfacing risks — it’s rebuilding how security services are