CISA Updates Voluntary Cybersecurity Performance Goals For Critical Infrastructure

The Cybersecurity and Infrastructure Security Agency has released an updated set of voluntary Cybersecurity Performance Goals for critical infrastructure operators, introducing more prescriptive and measurable controls aligned with current NIST guidance and the most prevalent threat vectors, including ransomware, phishing, and third‑party compromise. This update materially raises the bar for baseline cyber hygiene in sectors such as health care, energy, and transportation by emphasizing governance, continuous monitoring, and outcome‑driven security metrics.

Strategic Alignment With Modern NIST Frameworks

The revised goals are explicitly aligned to the latest NIST cybersecurity standards and frameworks, including updated control baselines for identity, asset management, data protection, and incident response. This alignment allows organizations that already use NIST Cybersecurity Framework or NIST SP 800‑series controls to map the CISA goals directly onto existing programs, reducing duplication and easing adoption. The goals emphasize outcomes such as demonstrable reduction in attack surface and time to detect and contain incidents, rather than prescribing only specific technologies.

Sector‑Agnostic Core Goals With Sector‑Specific Emphasis

While the goals are intended to be broadly applicable across critical infrastructure, CISA highlights sector priorities, with particular detail for health care, where legacy systems, constrained downtime windows, and sensitive personal and clinical data create elevated risk. Core expectations include comprehensive asset inventories, segmented networks for operational technology, and hardened external‑facing services. Sector overlays add requirements for protecting life‑safety systems, medical devices, and safety‑critical industrial control systems in environments such as hospitals and utilities.

Governance, Accountability, And Board‑Level Oversight

A major thematic shift in the updated goals is the emphasis on cyber governance and accountability at senior leadership levels. Organizations are expected to establish clear roles and responsibilities for cybersecurity, with board or executive oversight of risk posture, resourcing, and incident readiness. The goals describe governance practices such as risk‑based budgeting, formal risk acceptance processes, and periodic board‑level review of material cyber risks. This approach treats cybersecurity as an enterprise risk function tightly integrated with business and operational decision‑making, rather than a purely technical concern.

Measurable, Actionable Security Outcomes

The updated goals are framed as discrete, measurable actions rather than aspirational principles. Examples include targets for multi‑factor authentication coverage across privileged accounts, patching timelines for high‑severity vulnerabilities, and documented backup and recovery procedures tested to support specific recovery time and recovery point objectives. This outcome‑oriented design allows organizations to build internal scorecards, track progress over time, and demonstrate due diligence to regulators, insurers, and customers.

Emphasis On Common And High‑Impact Threats

CISA’s update prioritizes defenses against the most common and damaging threat types affecting critical infrastructure, such as phishing‑enabled credential theft, ransomware campaigns, exploitation of unpatched internet‑facing services, and compromises of managed service providers and software supply chains. Controls include widespread deployment of phishing‑resistant authentication, hardening of remote access pathways, active vulnerability management programs, and contractual security requirements for third‑party vendors. The goal is to ensure that limited resources are directed at threats that account for the majority of real‑world incidents.

Integration With Day‑To‑Day Operations

Rather than positioning cybersecurity as an overlay to operations, the updated goals stress integration into daily workflows, change management, and operational decision processes. This includes embedding security reviews into procurement, system design, and maintenance planning, as well as aligning cyber incident response with broader business continuity and crisis management plans. For operational technology and clinical environments, the guidance encourages coordination between engineering, clinical, and security teams to balance safety, availability, and security requirements.

Implications For Health Care And Other High‑Risk Sectors

In health care, the goals underscore the need to protect electronic health records, connected medical devices, and hospital operational systems from disruptions that could directly impact patient care. Recommended practices include network segmentation for medical devices, application of secure configuration baselines to clinical systems, and robust incident response runbooks tailored to clinical workflows. Other sectors such as water, transportation, and energy are expected to adapt the same outcome‑focused approach, prioritizing visibility into assets, resilient operations under cyber stress, and rapid containment of attacks targeting mission‑critical services.

SAP Releases December 2025 Security Updates Addressing Critical Vulnerabilities

SAP has published its December 2025 security updates, issuing fourteen new security notes that include three critical vulnerabilities affecting Solution Manager, Commerce Cloud, and the jConnect SDK. These flaws expose enterprises to risks such as remote code execution and data compromise in some of the most widely deployed components of the SAP ecosystem, making timely patching and compensating controls a high priority for SAP customers.

Scope And Severity Of The December Patch Set

The December release introduces a collection of new and updated security notes covering a range of products in the SAP portfolio. Among these, three issues are classified as critical, reflecting high exploitation potential and severe business impact if left unaddressed. The updated notes also revise prior advisories with additional technical details and remediation guidance, signaling that some earlier mitigations may require refinement or more complete deployment.

Critical Vulnerability In SAP Solution Manager

One of the critical issues affects SAP Solution Manager, the central management platform often used to administer and monitor large SAP landscapes. Because Solution Manager frequently holds elevated privileges and broad connectivity to other SAP systems, a remote code execution or privilege escalation flaw in this component can serve as an effective pivot point for attackers. Exploitation could enable an adversary to execute arbitrary code on the Solution Manager host, exfiltrate configuration and credential data, or issue malicious administrative commands against connected systems.

Commerce Cloud Exposure To Remote Exploitation

A second critical vulnerability targets SAP Commerce Cloud, the platform that powers many high‑volume e‑commerce deployments. In a typical architecture, Commerce Cloud systems process sensitive customer information and payment‑related data, and are exposed to the public internet. A critical flaw in this layer can enable compromise via crafted requests from unauthenticated or low‑privilege users, potentially leading to data theft, manipulation of catalog or pricing data, or disruption of order processing. Given the direct revenue impact of downtime or fraud in these environments, rapid application of patches or mitigating controls is essential.

jConnect SDK Risk In Upstream Integration Layers

The third critical issue resides in the jConnect SDK, a Java‑based connectivity toolkit widely used for integrating SAP databases and services with custom or third‑party applications. Vulnerabilities in SDK components are particularly problematic because they propagate through multiple consuming applications and services. A flaw in jConnect may allow attackers to leverage crafted connection parameters or protocol interactions to trigger memory corruption, injection, or unauthorized operations in applications that embed the SDK, expanding the blast radius beyond a single SAP instance.

Attack Surface Characteristics And Likely Exploitation Paths

From an attack surface perspective, these vulnerabilities affect both administrative backplanes and externally reachable services. Solution Manager instances often expose management interfaces and agents that, if left accessible from less trusted networks, can be probed for exploitable conditions. Commerce Cloud deployments are typically fronted by web application firewalls and content delivery networks, but misconfigurations or insufficient rule sets may permit exploit traffic to reach vulnerable endpoints. For jConnect, the primary risk lies in internal or partner‑facing applications that use the SDK for database connectivity, where adversaries may craft malicious inputs or leverage compromised application servers to trigger the weakness.

Mitigation, Patching Strategy, And Compensating Controls

Organizations operating SAP environments should prioritize applying the December security notes within an expedited change window, particularly for internet‑exposed Commerce Cloud instances and highly privileged Solution Manager systems. Where immediate patching is constrained by change‑control or testing requirements, compensating measures include restricting network access to management and integration interfaces, tightening authentication and authorization policies, and enabling detailed logging and monitoring for anomalous activity associated with the affected components. Segmentation between SAP management, application, and database tiers can further reduce lateral movement opportunities if a single component is compromised.

Broader Lessons For SAP Ecosystem Security

The December updates highlight the continuing importance of structured vulnerability management in large enterprise application stacks. Centralized management tools, public‑facing transactional platforms, and integration libraries all represent high‑value targets due to their connectivity and data access. Adopting a consistent process for rapidly ingesting vendor advisories, mapping them to specific system inventories, and implementing prioritized remediation helps reduce exposure windows. In parallel, architectural defenses such as least‑privilege administration, rigorous authentication, and layered network controls remain critical to limit the impact of any newly discovered SAP vulnerabilities.

React Issues Additional Security Patches Following New Vulnerability Discoveries

The React ecosystem has received a new round of security patches after researchers identified additional flaws affecting deployments of the framework, including sites operated by critical infrastructure providers and government entities. These vulnerabilities underscore how client‑side JavaScript frameworks, when combined with complex supply chains and third‑party integrations, can become vectors for cross‑site scripting, injection, and data exfiltration attacks at significant scale.

Nature Of The Newly Addressed Vulnerabilities

The latest patches address issues in the React framework and related tooling that could allow attackers to manipulate component rendering or bypass existing protections against script injection. In some cases, improper handling of input, encoding, or component props can enable injection of untrusted content into the Document Object Model, reintroducing classes of cross‑site scripting vulnerabilities that developers often assume are mitigated by default. Additional concerns relate to ecosystem packages and build configurations that extend React, which can introduce unsafe behaviors if not carefully audited.

Impact On Critical Infrastructure And Government Sites

Security researchers have warned that some of the vulnerable React deployments belong to organizations in critical infrastructure sectors and public‑sector agencies, where web portals handle sensitive operational or citizen data. Successful exploitation on these sites could facilitate credential theft, session hijacking, or unauthorized access to administrative functions. In environments where web front ends are integrated with operational technology dashboards or case management systems, a compromised browser session could provide attackers with indirect leverage over underlying control systems or systems of record.

Technical Characteristics Of The Attack Surface

React applications typically rely on a complex build and packaging toolchain involving bundlers, transpilers, and dependency managers. This stack expands the attack surface, as vulnerabilities may arise not only in React core but also in JSX transformation, server‑side rendering components, or libraries that handle routing, state management, and data fetching. An attacker may target weak validation in API responses that are rendered into components, exploit unsafe use of dangerouslySetInnerHTML, or abuse prototype pollution in supporting libraries to influence runtime behavior of React components and inject malicious code paths.

Patch Deployment And Secure Configuration Practices

Applying the newly issued patches involves upgrading React and relevant ecosystem packages to the latest secure versions, followed by regression testing of application behavior. Organizations should review their use of any mechanisms that render raw HTML or dynamic script content, ensuring that sanitization and encoding are applied consistently and that untrusted data does not flow directly into rendering sinks. Build configurations should be examined to disable unsafe optimizations, enforce strict content security policies, and minimize the inclusion of unvetted third‑party scripts or components.

Defensive Monitoring And Runtime Protections

Beyond patching, defenders can strengthen monitoring around React‑based front ends by instrumenting client‑side error logging, anomaly detection for script behavior, and inspection of network calls made by the application. On the server side, logging of API usage, unusual parameter patterns, and authentication anomalies associated with React front‑end sessions can reveal attempted exploitation. Where possible, runtime application self‑protection and modern web application firewalls tuned to React application behavior can help identify and block payloads that aim to exploit weaknesses in component rendering or state handling.

Implications For JavaScript Framework Security Practices

The latest round of React patches illustrates that reliance on a popular framework does not eliminate the need for secure coding, dependency hygiene, and architectural defenses. Development teams should treat framework updates as part of continuous security maintenance, integrating dependency scanning and automated alerts for new advisories into their pipelines. Regular review of third‑party packages, least‑privilege principles in API design, and defense in depth at the browser, application, and API layers remain essential to mitigate the evolving threat landscape for modern JavaScript applications.

OWASP Publishes Inaugural Top 10 Risks For Agentic AI Applications

The Open Worldwide Application Security Project has released its first Top 10 list focused on agentic AI applications, providing a structured taxonomy of the most critical security risks associated with autonomous AI agents that can plan and execute tasks with limited human oversight. This framework gives organizations a starting point for systematically analyzing and mitigating novel threats such as goal hijacking, unsafe tool use, and emergent behaviors in complex AI workflows.

Defining Agentic AI And Its Security Challenges

Agentic AI refers to systems that couple large language models or other AI components with tools, memory, and decision logic that allow them to autonomously decompose objectives into tasks, interact with external systems, and iteratively refine their actions. Unlike traditional static models, these agents maintain state over time and can invoke tools such as databases, code execution environments, or external APIs. This expanded capability surface introduces new failure modes, including unintended actions, environmental manipulation, and adversarial influence over goal formation or tool invocation sequences.

Core Risk Categories In The New Top 10

The Top 10 for agentic applications organizes risks into categories that reflect both classical security issues and AI‑specific concerns. Examples include manipulation of an agent’s goals or instructions, unauthorized or unsafe tool usage, data exfiltration through indirect prompt or tool interactions, and inadequate oversight or guardrails around autonomous decision loops. Additional categories address weaknesses in identity, authentication, and authorization for agents themselves, as well as supply‑chain risks in the tools and plugins they rely on.

Goal Hijacking And Instruction Manipulation

Goal hijacking occurs when an attacker influences the objectives or internal state of an agent, causing it to pursue actions that benefit the adversary rather than the intended user. This can be achieved via adversarial prompts, poisoned data sources, or compromised tools that return crafted outputs designed to alter the agent’s plan. Because agentic systems often re‑plan and iterate based on intermediate results, a single manipulated step can cascade into a series of harmful actions, such as modifying infrastructure configurations, exfiltrating sensitive data, or initiating fraudulent transactions.

Unsafe Tool Use And Environment Interaction

Many agentic AI systems are granted access to powerful tools, including code execution environments, shell interfaces, database clients, and cloud management APIs. The Top 10 highlights the risk that agents may invoke these tools in unsafe ways if constraints, validation, and policy enforcement are insufficient. For instance, a coding agent might execute untrusted input as code, delete or misconfigure resources while attempting remediation, or generate infrastructure changes that violate security baselines. The combination of tool access and partial autonomy magnifies the potential impact of both benign errors and adversarial interference.

Data Security, Privacy, And Supply‑Chain Risks

Agentic applications frequently aggregate and process data from multiple systems of record, third‑party services, and user inputs, raising the risk of unintentional data leakage across contexts. The Top 10 addresses scenarios where agents expose sensitive information through responses, logs, or tool outputs in ways not anticipated by designers. It also calls attention to the supply‑chain surface created by tool plugins, external connectors, and model hosting services. Compromised plugins or model endpoints can provide adversaries with a foothold to manipulate agent behavior or access downstream systems trusted by the agent.

Recommended Mitigation Strategies

To address these risks, OWASP recommends a layered approach that includes explicit policy and constraint definition for agent actions, robust authentication and authorization for all tools and data sources, and comprehensive logging of agent decisions and tool invocations for forensic analysis. Additional practices include sandboxing high‑risk tools, limiting blast radius through fine‑grained permissions, and performing adversarial testing of agent behavior under crafted inputs. Governance processes should treat agentic AI systems as high‑privilege software components, subject to threat modeling, change control, and continuous monitoring similar to other critical applications.

Implications For Secure AI Adoption

The publication of a dedicated Top 10 for agentic applications provides security, engineering, and risk teams with a common vocabulary to evaluate AI‑driven workflows. By integrating these risk categories into existing application security and threat modeling practices, organizations can avoid treating AI agents as opaque black boxes and instead subject them to structured scrutiny. This enables more informed decisions about where and how to deploy autonomous capabilities, how much operational authority to delegate to agents, and what safeguards are necessary to maintain control and accountability in AI‑augmented systems.