Superna Data Security for Dell AI Infrastructure
Know who touched your AI data, and whether it can still be trusted
AI workloads on Kubernetes spin up, read sensitive data, and disappear before anyone reviews what they did. Superna links every Dell PowerScale and ObjectScale operation to the pod, the job, and the person behind it, classifies sensitive data as it lands, and scores content integrity to catch tampering, poisoning signals, and other threats before they reach production AI.
Provenance
Where did the data in your training set or RAG store come from, and who put it there? Without a record, unreviewed data can enter an AI pipeline unnoticed.
Access
Storage audit shows a service account, and Kubernetes keeps no data-level audit. When an auditor asks who touched a regulated dataset behind an AI application, the answer is buried across silos.
Manipulation
Training data, retrieval stores, and model artifacts can be altered, encrypted in place, or corrupted without triggering an access alert. The damage surfaces later, when models or answers degrade.
Exfiltration
A throwaway pod can mount a sensitive volume, copy data out, and be deleted before anyone looks. Off-hours bulk reads blend into normal volume.
01
See
Map every user, pod, PVC, bucket, and storage volume in your AI pipeline, with sensitive-data classification overlaid. Rewind the map for historical views.
02
Attribute
Resolve every file and object event to the pod, the job, and the person or automation identity behind it, even after the pod has been deleted.
03
Detect
Surface the changes that matter, such as files becoming sensitive, collapsing content integrity, short-lived pods reading sensitive volumes, and off-hours throughput spikes. Severity escalates only when multiple signals correlate.
04
Prove
Answer who accessed a regulated dataset with an attributed record, trend sensitive-data and integrity populations over time, and export forensic audit evidence.
Built for Kubernetes on Dell Storage
All it takes is Kubernetes on Dell storage
Superna Data Security for Dell AI Infrastructure requires only Kubernetes workloads running on Dell PowerScale or Dell ObjectScale. It works with or without Dell AI Factory and the Dell AI Data Platform, and gains added depth from NVIDIA Run:ai and the Dell Data Analytics Engine wherever you run them.
Dell PowerScale and ObjectScale
File and object audit, attributed to
pods and people
NVIDIA Run:ai (optional)
Names the data scientist
behind each GPU job
Dell Data Analytics Engine (optional)
Structured-data audit and behavioral anomaly detection
Identity Lineage for AI Workloads
Trace every PowerScale and ObjectScale event to a pod, a job, and a person. Distinguish the Pod Creator from the identity the platform used, so GitOps, CI, and scheduler-launched workloads still trace to a user.
Live Data-to-Workload Map
See which workloads mount which volumes and buckets, and which storage holds the data. Historical views allow you to time-travel the topology to see what existed and what was connected.
Sensitive-Data Discovery as Data Lands
Classify PII and other sensitive data as it is written into training, RAG, and shared volumes, not on a periodic scan cycle. Know which workloads and users reach it.
Content Integrity Scoring
Read whether AI data is still coherent. Detect in-place encryption, corruption, gradual integrity drift, and manipulation signals in training sets and retrieval stores that activity monitoring alone can miss.
Ephemeral and Deleted Workload Detection
Flag short-lived, deleted, and repeatedly recreated pods, PVCs, and buckets, along with who removed them. These patterns are often signs of exfiltration staging and attacker cleanup.
Structured Data Audit
For the Dell Data Analytics Engine or Trino, record who ran which SQL against which tables. Baseline each user’s behavior and flag bulk export, scope expansion, destructive operations, and probing.
AI data trust across regulated industries
Wherever regulated or high-value data enters AI pipelines, security and governance teams need to prove who touched it and whether it can be trusted.
Healthcare and Life Sciences
See where patient data lands in AI datasets, which workloads read it, and who launched them.
Financial Services
Produce attributed access records for data behind financial AI models, and detect insider activity across file, object, and SQL data.
Manufacturing
Attribute every change to production AI training data and engineering datasets to a named user and workload.
Research and Higher Education
Show who added, changed, or removed data in shared research datasets before it shapes a model.
WORKS BEST WITH
Pair with Superna Data Security Edition. Data Security Edition protects and recovers your storage estate. Superna for Dell AI Infrastructure proves who touched the AI data on it and whether it is intact.
Ready to close the gap?
Over 3,600 organizations rely on Superna to defend their data where it’s most exposed. Let’s talk about yours.
Take a Tour Watch the video
