Talastron is a signatory of the Armed Forces Covenant.
Talastron has signed the Armed Forces Covenant, a pledge to support the armed forces community and treat those who serve, or have served, with fairness and respect.
What we hold, what we're actively pursuing, and where the architecture is aligned. Each one dated and sourced. Nothing else.
Talastron has signed the Armed Forces Covenant, a pledge to support the armed forces community and treat those who serve, or have served, with fairness and respect.
Talastron is certified under the UK government-backed Cyber Essentials scheme, demonstrating baseline controls against common cyber threats. The certificate is held under the company’s prior registered name, Orion Data Analytics Ltd, now Talastron Ltd.
Talastron is registered with the UK Information Commissioner’s Office under the Data Protection Act, as required of organisations that process personal data.
Participation in the Microsoft ISV Success Program supports Talastron Kinetic AI’s engineering and publication path towards the Azure Marketplace.
Talastron holds active Microsoft Partner status (Partner ID 7100919), in the Microsoft AI Cloud Partner Program and the ISV Success Program. This underpins Talastron Kinetic AI’s delivery model: manufactured software running inside the customer’s own Microsoft Azure tenant.
Talastron Kinetic AI’s gated, governed pipeline maps its controls to ISO/IEC 27001, and certification work is under way against that mapping. This is a pursuit-in-progress statement — it will move to “Achieved” once the certificate is issued, not before.
Every regulatory, partnership and certification claim on the site is factually current and reviewed before publication. 'Achieved' is reserved for credentials we hold; 'In progress' means we're actively pursuing it, with work under way; 'Aligned to' describes how the architecture is built, with no certification claim attached. We won't call anything ISO 27001 certified until it is.
AI Trust, Risk & Security Management · AI TRiSM
AI TRiSM is Gartner's framework for ensuring that agentic AI systems operate reliably, securely, fairly, and in compliance with data-privacy regulations. The Five-Gate Manufacturing Model implements all four pillars at defined, human-approved gates — not as an afterthought, but as the production process itself.
Trust
Every AI output is adversarially reviewed by Themis before it reaches a human gate. The Dual-Judge pattern — Primary Judge then Formatting Judge — ensures outputs are accurate, complete, and interpretable before any human is asked to approve them. The WAF scorecard at Gate 5 is a live readiness assessment of the deployed system, not the design.
Operations
The Five-Gate pipeline enforces end-to-end lifecycle governance on every agent output. Gate 3 applies ISO 42001 AI governance controls — explainability, bias assessment, model risk — before any infrastructure is built. Azure ML model lineage is captured and transferred to the client at Gate 5 as part of the IP handover.
Security
Vulcan generates AVM-first Bicep with zero hardcoded secrets, no public endpoints on data or AI services, and tagging lock on every resource. Cassandra cross-checks every predicted deployment change against Nemesis's live Azure Policy constraints before Gate 4 opens. No resource reaches the client's tenant without policy verification.
Risk
All data remains inside the client's Azure tenant — zero egress, UK South sovereign. Gate 1 classifies PII, IP, and regulated data before architecture is designed. UK GDPR data minimisation and purpose limitation are applied at Gates 1 and 3. Consumer Duty explainability is verified at Gate 5 for regulated financial deployments.
How Talastron references AI TRiSM in analyst briefings: The Five-Gate pipeline wraps probabilistic LLM reasoning in deterministic validation contracts, context-isolated agent boundaries, and automated DevSecOps policy checks before any code or action reaches enterprise production systems. This is Talastron's implementation of the AI TRiSM framework.
Bring a real requirement. We build a working architecture together and you walk out with a spec and a fixed-price commitment.