Every story tagged Vulnerability Disclosure, curated for CIOs and IT leaders — ranked by source credibility, engagement, and freshness.
823 stories · open in the command center
A now-patched flaw in AWS Bedrock AgentCore shows how a single malicious prompt can turn an AI agent into a foothold for stealing temporary AWS credentials, accessing secrets, and potentially taking over all agents in an account and region. For CIOs, the strategic takeaway is that agentic AI can dramatically expand the blast radius of cloud misconfigurations, making least privilege, network isolation, and tight control over agent permissions essential to keeping AI from becoming a new control-plane risk. IT organizations should treat AI agents like privileged infrastructure components, not simple apps, and validate that vendor defaults do not create cross-agent lateral movement or secrets exposure.
The OMGF | GDPR/DSGVO Compliant, Faster Google Fonts. Easy. plugin for WordPress is vulnerable to Stored Cross-Site Scripting via 's' Search Parameter via comments-atom Feed in all versions up to, and including, 6.3.10 due to insufficient input sanitization and output escaping. This makes it possible for unauthenticated attackers to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. Successful exploitation requires that the front-end server serves the retained .tmp file without a Content-Type or X-Content-Type-Options header, enabling MIME-sniffing browsers such as Chromium to execute the injected script — a condition present by default on many Apache and nginx/php-fpm deployments.
Path traversal vulnerability in the Satel Iberia SenNet Datalogger Serie 200, specifically in the web portal provided by the device, which allows an authenticated user to read any file or list any directory accessible to the system user running the web server. This is possible by modifying the URL to include a path traversal payload. Successful exploitation of this vulnerability could allow an attacker to access critical system files containing confidential information.
Path traversal vulnerability in the BugTracker.NET file download component. The parameter used to specify the file name does not properly validate user-supplied paths. An authenticated remote attacker could enter a manipulated path to access files located outside the intended directory. Successful exploitation could allow the attacker to read system files accessible to the account used by the application.
Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection') vulnerability in Gopiplus Post title marquee scroll post-title-marquee-scroll allows Blind SQL Injection.This issue affects Post title marquee scroll: from n/a through 9.9.
Rundeck before 6.2.0 contains an OS command injection vulnerability that allows authenticated users with job run permission to execute commands on Windows nodes by supplying crafted option values. Attackers can inject cmd.exe metacharacters such as && or | into free-text options, which CLIUtils.quoteWindowsCMDArg wraps in ineffective single quotes, running commands with node executor credential privileges.
MiniUPnPd through 2.3.11 built with --strict contains a divide-by-zero vulnerability in ProcessSSDPData() that allows unauthenticated local network attackers to crash the daemon. Attackers can send a single multicast M-SEARCH datagram with MX: 0 and a known ST to port 1900, triggering SIGFPE and denying UPnP IGD service.
llama.cpp before b11393 contains a use-after-free and double free vulnerability in common_chat_peg_mapper::map that allows unauthenticated remote attackers to corrupt heap memory via a dangling current_tool pointer. Attackers can submit a chat_parser in a POST /completion request emitting a tool-id after a tool-close tag to crash llama-server and shape a heap write primitive.
A command injection vulnerability exists in the h-ui (version v0.0.25 and below) administrative API due to improper validation of the listen configuration field. When an authenticated administrator submits a value containing shell metacharacters, the application constructs nftables/iptables rule strings using fmt.Sprintf and executes them via bash -c as root. Because the listen field lacks port or format validation, arbitrary OS commands can be injected and executed with root privileges.
libexpat before commit 13c5f63 contains a heap buffer over-read vulnerability in xmlparse.c. XML_ParseBuffer advances the parse buffer end with parser->m_bufferEnd += len using a caller-supplied length that is not validated against the allocated buffer size, so repeated XML_ParseBuffer calls move m_bufferEnd past the end of the heap allocation and subsequent parsing reads out of bounds. Reaching this path requires a parse buffer to already be present; otherwise XML_ParseBuffer returns XML_ERROR_NO_BUFFER. A buffer is present after a prior call to XML_GetBuffer, either directly (the common case) or indirectly through a prior XML_Parse call that allocates the buffer internally. The over-read discloses adjacent heap memory to the calling application, recovering heap pointers, libc function pointers, and code pointers sufficient to defeat ASLR and build further exploitation primitives.
LMCache through 0.5.5 contains a missing authentication vulnerability in the multiprocess mode HTTP server that allows remote unauthenticated attackers to access management endpoints listening on all interfaces by default. Attackers can read environment credentials via GET /env and configuration via GET /config, clear caches, delete cache objects, and modify tenant quotas to evict other tenants' cached data.
LMCache through 0.5.5 contains a server-side request forgery vulnerability in its frontend monitoring service that allows unauthenticated attackers to bypass the proxy allowlist by registering arbitrary hosts. Attackers can add entries via POST /api/proxies and then use /proxy or /proxy2 to reach internal hosts, read responses, and tamper with nodes or stop the heartbeat.
: Session Fixation / Session Reuse across Users vulnerability in Apache Jackrabbit. Jackrabbit WebDAV server attaches a cached authenticated session on any Lock-Token/TransactionId/SubscriptionId/If-header field token match with no credential check. This issue affects Apache Jackrabbit: from 2.23.0 through 2.23.5, from 2.22.0 through 2.22.4, from 2.20.0 through 2.20.17. Users are recommended to upgrade to versions 2.23.6, 2.22.5, or 2.20.18 which fix the issue.
As part of Cisco's ongoing commitment to proactive security and product quality, the Cisco NX-OS engineering team has conducted a comprehensive internal security review. This review resulted in a software hardening release that addresses multiple internally discovered vulnerabilities. The vulnerabilities tracked by CVE-2026-76455 are related to improper access control issues that are grouped under the Common Weakness Enumeration (CWE) CWE-284.
Atlassian’s newly patched arbitrary file access flaw (CVE-2026-21589) is already being actively scanned in the wild, which raises the likelihood of rapid exploitation against internet-facing Jira, Bitbucket, and Confluence instances. For CIOs and IT leaders, the business risk is exposure of sensitive configuration and application data that could lead to broader compromise, making patch velocity, asset visibility, and log-based threat hunting immediately important priorities for enterprise security teams.
Use after free in Fonts in Google Chrome on on Windows prior to 155.0.8059.39 allowed a remote attacker to execute arbitrary code inside the sandbox via a crafted HTML page. (Chromium security severity: Medium)
Incorrect authorization in Selection in Google Chrome on on iOS prior to 155.0.8059.39 allowed a remote attacker to obtain sensitive information via a crafted HTML page. (Chromium security severity: Low)
Information leak in Passwords in Google Chrome prior to 155.0.8059.39 allowed a remote attacker to obtain sensitive information via a crafted HTML page. (Chromium security severity: Low)
Integer overflow in Media in Google Chrome prior to 155.0.8059.39 allowed a remote attacker to potentially execute arbitrary code outside the sandbox via a crafted HTML page. (Chromium security severity: Low)
Incorrect authorization in Browser in Google Chrome prior to 155.0.8059.39 allowed a remote attacker to obtain sensitive information via a crafted HTML page. (Chromium security severity: Low)
The CakeResponse::download() method in lib/Cake/Network/CakeResponse.php constructs a Content-Disposition header by directly interpolating a caller-supplied filename into a quoted-string value without sanitization. Two distinct injection vectors exist in the unpatched code. First, if the filename contains C0 control characters (CR or LF), PHP refuses to emit the entire Content-Disposition header, silently dropping the attachment disposition. The response body is then served with its own Content-Type (for example text/html for an .html attachment) and renders inline in the browser on the application origin, creating a stored cross-site scripting condition. The commit message notes this is reachable even when the download_attachments_on_load setting is enabled, meaning a victim merely needs to view a page that triggers the download. Second, a double-quote character in the filename terminates the quoted-string value early, permitting injection of additional Content-Disposition paramete...
Hydra is a framework for elegantly configuring complex applications. From 1.2.0 until 1.3.0 and 1.4.0.dev10, the hydra-optuna-sweeper package accepts a configuration-controlled dotted path in hydra.sweeper.custom_search_space, resolves it with hydra.utils.get_method(), and later invokes the returned callable in the Hydra controller process. Because get_method() is a trusted-input lookup helper and does not apply the execution policy used by instantiate(), an attacker who controls Optuna sweep configuration or command-line overrides can select importable Python code for execution with the application's privileges, including bypassing a trusted execution whitelist on affected Hydra 1.4 development releases. This issue is fixed in versions 1.3.0 and 1.4.0.dev10.
Memory corruption vulnerabilities exist in AOS-S that are reachable by an unauthenticated adjacent attacker. Successful exploitation could allow an attacker to execute arbitrary code.
A missing integrity verification vulnerability exists in the OnGuard agent of ClearPass Policy Manager. Successful exploitation could allow an unauthenticated, remote attacker to execute arbitrary code on the affected endpoint with the elevated privileges of the agent.
A format string vulnerability in an affected service interface of HPE Networking ClearPass Policy Manager could allow an unauthenticated remote attacker to corrupt process memory. Successful exploitation could allow an attacker to execute arbitrary code.
SQL injection vulnerabilities in the web-based management interface of ClearPass Policy Manager could allow a low-privileged authenticated remote attacker to conduct SQL injection attacks against the ClearPass Policy Manager instance. Successful exploitation could allow an attacker to run arbitrary database commands.
A missing integrity verification vulnerability in the Windows client software for ClearPass Policy Manager could allow malicious users on a local instance to elevate their user privileges. A successful exploit could allow these users to execute attacker-supplied code with elevated privileges on the local system.
An unauthenticated path traversal vulnerability exists in an API endpoint of ClearPass Policy Manager. Successful exploitation of this vulnerability allows an unauthenticated remote attacker to influence authorization decisions and be assigned an unintended role.
Authorization Bypass Through User-Controlled Key (CWE-639) in Kibana could lead to cross-tenant data interception. In this context, "tenant" refers to a user or team sharing the same Kibana deployment, not a separate Elastic Cloud organization or customer. Kibana's Fleet package installation process allowed a user holding delegated Fleet package-management privileges, without direct Elasticsearch administrative privileges, to claim a data stream identifier already in use by another tenant. Because ownership of that identifier was not verified before Fleet applied the uploaded package's generated index and ingest-pipeline settings to already-existing infrastructure, an attacker could redirect an existing tenant's data stream through infrastructure under their control. This exposed the affected tenant's subsequently ingested data to unauthorized disclosure and modification, and prevented that data from reaching its intended destination. Interception could continue even after the malic...
Anthropic has merged its security access programs into a three-tier model that gives different levels of Claude capability to defense teams, red teams, and highly trusted critical-infrastructure organizations. For CIOs and technology leaders, the strategic takeaway is that AI is quickly becoming embedded in security testing and vulnerability discovery, but the operational bottleneck remains remediation: Anthropic says its programs have surfaced more than 129,000 verified vulnerabilities while only a fraction have been patched. IT organizations should expect AI-assisted security tooling to raise the volume and speed of findings, requiring tighter vulnerability management, patch prioritization, and governance around model access and data retention.