Your data, under your control.
The detail security and procurement teams need: how Blacklight is certified, how your data is isolated and where it lives, how we ensure it is never used to train AI models, and how it is permanently deleted when you leave. For the shorter version, see the FAQ.
Independently certified.
Formal security controls span access management, encryption, secure development, vendor risk, logging and incident response across the platform. Further attestations are on our roadmap; the current certification and our full control set are available on request.
Your data is never used to train AI models.
No customer data is used to train or improve Blacklight’s models or any third-party model. That commitment is enforced at three levels.
Every third-party model provider is engaged under an enterprise contract that explicitly prohibits using customer data for training, fine-tuning, evaluation or human review. Any new provider passes a contractual review confirming the same no-training position before integration is approved.
Inference-only access: your data is only ever sent to a provider’s inference API, never a training, fine-tuning, RLHF-feedback or model-improvement endpoint. No customer data is used to fine-tune any model. Zero-retention is enabled with every provider that offers it, so prompts and completions are not persisted beyond the inference call.
Annual review of all provider contracts and policy updates to confirm the no-training commitments remain in force. A vendor security questionnaire and Data Processing Agreement are held on file for each provider, and any change in a provider’s data-handling posture is assessed under our Vendor Risk Management programme before continued use.
Proprietary by design.
The specific identity of the third-party AI models and providers Blacklight uses is treated as commercially confidential. The composition of the model stack, the terms governing each relationship, and the deployment architecture are material proprietary intellectual property and are not disclosed in open questionnaires or public documentation. The position is commercial rather than evasive, and is fully consistent with your right to assurance over how your data is handled: the protections above apply regardless of which models are involved.
Your own instance, and in your region.
Every client runs on a dedicated, single-tenant instance. Not a shared platform with logical separation between customers: your telemetry, alerts and case data live in your own instance, and are never co-mingled with, or reachable from, another client's. Each instance is encrypted with keys unique to you. Only the minimum context required for a given inference call ever leaves that boundary, with sensitive identifiers (PII, credentials, internal IPs and hostnames) redacted or tokenised in transit. Access is least-privilege and audited, and data is encrypted in transit and at rest.
Running a managed service? MSSPs and partners get a multi-tenant console to operate every client from one interface. That console governs your team's visibility only: each client remains a separate tenant with separately segregated data, and access is scoped by role to the tenants a given analyst is authorised to see.
Blacklight is hosted across multiple regions, globally and locally. You choose your hosting region at contract from the regions we support, and your data at rest stays there, with in-country deployment available where we operate. That matters for regulated and sovereign environments. Tell us the jurisdiction you need and we will confirm what is available.
When you leave, your data leaves with you.
On contract termination or service exit, Blacklight runs a documented deletion process that covers every system in which your data may have been stored, processed or cached, and produces an auditable record of completion.
- Primary stores SIEM telemetry indices, case data, alerts, investigation artefacts, embeddings and derived ML features.
- Replicas & read paths Secondary databases, search indices and analytics warehouses.
- Caches & ephemeral state In-memory caches, message queues, CDN edges and agent working memory.
- Backups All snapshot and point-in-time backup tiers, rolled off at the date of effective termination.
- Third-party model context Any retained API context is purged in line with each provider’s deletion API or natural retention window (zero-retention is the default configuration).
Records are marked for deletion at termination and physically removed by automated purge jobs. Encrypted-at-rest data is crypto-shredded: destroying the per-tenant Data Encryption Key renders the underlying ciphertext mathematically irrecoverable, even where backup media has not yet been overwritten. Decommissioned physical media follows NIST SP 800-88 Rev. 1.
On request, Blacklight issues a certificate of completion confirming deletion, listing the systems and data in scope, the deletion method applied, and the completion date. Your data is available for export for 30 days from termination before deletion runs.
Need our full security pack?
We can share the ISO 27001 certificate, our Data Processing Agreement, and answers to your security questionnaire under NDA. Ask and a real person replies.
See what truly predictive
security looks like.
Sixty minutes, under your control. See the platform run on a pre-loaded, anonymised environment, watch one real incident handled end to end, and map it to your sector, without connecting a single data source.
- 01 A live tour on a pre-loaded, anonymised environment, running from minute one, not slides.
- 02 One real incident, detected, investigated and contained, written up as a regulator-ready report.
- 03 Mapped to your world: your sources, your sector's threats and your regulators.
- 04 The questions your board will ask: deployment, residency, security, integrations and TCO.
No connectors or data required from you. A proof-of-value on your own telemetry is the next step, never the ask on a first call.