Our Container Security AMA: You asked, Wiz answered

Check out the top comments and responses from our recent containers AMA.

Last week, we hopped on the /cybersecurity subreddit for a container security AMA and wow, what great questions from the community! We talked Kubernetes, cloud misconfigurations, AI, and IAM. It was clear that as more and more workloads move to Kubernetes and cloud, security practitioners must evolve their approach and rethink how they secure these ephemeral, decentralized environments in a variety of ways.

Here’s a recap of the best questions, hottest takes, and key points from the AMA. You can viewthe full discussion on Reddit.

Our answer: having 100% container image security coverage. Container images often contain vulnerabilities inherited from third-party dependencies, so teams must ensure that only trusted, verified images are deployed.

Wiz recommends a multi-layered approach here. All container images should be scanned before deployment to detect CVEs, malware, and exposed secrets. Organizations should enforce deployment controls by using signed and verified images or restricting deployments to pre-approved registries with a deny-all default policy. Lastly, security must be integrated throughout the SDLC, covering CI/CD pipelines, deployment, and runtime with solutions like Cloud Workload Protection Platforms (CWPP) and Extended Detection and Response (XDR) tools to ensure continuous protection.

The topic of container security challenges was a recurring theme throughout the AMA, and a few other points bubbled up, too:

IAM: Managing permissions and identities in containerized environments is difficult, particularly in Kubernetes, where misconfigured RBAC settings can lead to security gaps.

Runtime Security & Detection:Monitoring container workloads in real time is complex due to their ephemeral nature. Attackers can exploit misconfigurations before detection tools catch them. Check out ourBSides San Franciscotalk on Effective Detection in Kubernetes Clusters for more info.

K8s Control Plane Security:Many organizations struggle to secure their Kubernetes control plane, especially with cloud-managed clusters where visibility is limited.

Multi-Tenancy:In shared Kubernetes environments, namespace isolation weaknesses can lead to unauthorized access across different tenants. (Wiz builtNamespaceHoundtodetect these risks.)

Credit to ImaginaryWheatThins for this very timely question– it’s a topic that comes up every day for us at Wiz and probably anyone who is working in technology today.

Shay answered that Kubernetes is a big facilitator for many technologies and AI in particular, given the scale and sensitivity of AI workloads. He said that AI has "sharpened the existing Kubernetes Threat Model," pointing out that "the AI model IS an executable code" and flagging vendor multi-tenancy issues.

Zooming out, AI carries massive potential for innovation and disruption, but governance and security must keep pace in order to avoid incurring serious risk. Multiple discoveries from the Wiz Research team underscore this reality:

DeepLeak, an exposed DeepSeek database leaking sensitive information – including chat history and over a million lines of log streams – and allowing full control over database operations.(Note that we found this after the AMA’s completion.)

A critical NVIDIA AI vulnerabilitythat affected containers using NVIDIA GPUs, including over 35% of cloud environments. CVE-2024-0132 presented high risk to AI workloads and environments.

Our work with AI-as-a-service providersHugging FaceandReplicaterevealed that malicious models pose a major risk to AI systems and customer data in particular.

Probllama, an easy-to-exploit Remote Code Execution vulnerability in the open-source AI infrastructure project Ollama.

SAPwned, vulnerabilities in SAP AI Core that would allow malicious actors to take over the services and access customer data by exposing cloud environments and private AI artifacts.

Read more on this question.

We love that this came from a self-proclaimed happy Wiz customer! Their inquiry was about balancing deep visibility while minimizing performance impact.

Our response emphasized both agentless and agent-based coverage. The former is where Wiz has its roots, and how our Wiz Cloud offering matured into what is now a near real-time CSPM via agentless scanning. Agent-based coverage is how we offer real-time monitoring and threat detection, with agents designed for large-scale environments.

To elaborate a bit more on our AMA responses: for cloud-native workloads, organizations should adopt agentless, API-driven security tools to ensure seamless monitoring without impacting performance. For runtime protection and non-cloud-native workloads, CWPP, CDR, and Admission Controllers provide deeper security controls. Additionally, teams should put a premium on privacy and data protection by ensuring that sensitive data and personally identifiable information (PII) are properly redacted to prevent unauthorized exposure.

Read more on this question.

****

Thanks for having us, Reddit! It was a blast diving into container security challenges and exchanging insights with the community. These discussions are what make shared spaces so valuable—bringing security practitioners together to exchange knowledge, tackle tough problems, and level up collectively. We appreciate all the great questions and perspectives, and we’re already looking forward to the next one. Until next time, keep securing those containers, and see you around the subreddit! 🚀🔒

Discover new approaches to securing your containerized apps and Kubernetes clusters in this user-friendly book.

The Wiz Certified program refines your cloud security skills to help you grow your career and stand out among your industry peers.

Discover key strategies to strengthen your AWS security posture, from applying protection at all layers to understanding shared responsibility in the cloud.

A publicly accessible database belonging to DeepSeek allowed full control over database operations, including the ability to access internal data. The exposure includes over a million lines of log streams with highly sensitive information.

Get a personalized demo

©2026Wiz, Inc.

StatusPrivacy PolicyTerms of UseModern Slavery StatementCookie Settings