Security
Security, control and trust.
Independently audited to SOC 2 Type II standard and built entirely on Microsoft Azure. Your data is held in its own database, stored in the UK, and never used to train models.
SOC 2 Type II
Audited by A-LIGN, covering the Lynctera platform Almagest runs on. Request the report from our Trust Centre.
Data protection
GDPR-compliant data handling, with client data stored in the United Kingdom.
One cloud
Data, documents and AI models all run on Microsoft Azure under enterprise terms.
ISO/IEC 42001
Certification of our AI management system, the international standard for governing AI.
Data protection
How your data is protected.
Your own database
Each client’s data is held in its own dedicated database, and every request is tied to a single client before any data is read.
Encryption
AES-256 encryption at rest, including backups, and TLS 1.2 or higher in transit.
No training on client data
Your documents are never used to train models. Not ours, not anyone’s.
AI through Microsoft Azure
Models from OpenAI and Anthropic are reached only through Microsoft Azure, under enterprise terms that exclude training on your data. We never call the model companies’ own services.
Stored in the UK
Client data is stored in Microsoft Azure’s UK South region, with encrypted geo-redundant backups, point-in-time restore across the last 30 days and long-term backups kept for up to two years.
Controlled access
Staff access to the production environment runs through Microsoft Entra ID with multi-factor authentication and conditional access.
Controlled change
Prompts, models and tools are held in one governed store. Every change is recorded, can be compared with the version before it, and can be rolled back.
Scanned before release
Every code change is scanned for vulnerable dependencies and exposed secrets before it reaches production.
Auditability & traceability
Activity logs across users, agents, documents and outputs, kept for a year, and every decision traced back to source.
Control & trust
A system you can control, inspect and trust.
Almagest is built with a proprietary harness and context engineering. Agentic flows are tightly bounded: they follow the process your team defines, they show their working, and they stop where your judgement begins.
Nothing replaces your investment committee.
Controlled by design
Every workflow follows the steps, sources and sign-offs your team defines, on every run. The AI models inside it produce probabilistic results, which is why every output is cited and every decision stays with your team.
Explainable outputs
Every number, claim and conclusion carries a citation to the source document and page.
Auditable by construction
Every model step and every human sign-off writes a record automatically, aligned to ISO 42001.
Human in the loop
The platform prepares the evidence. Your investment team makes the decision.
Your oversight
Evidence for the people who supervise you.
Almagest is designed to support your compliance obligations. It keeps the evidence your oversight depends on: who decided what, on which sources, and when.
Personal data
Client data is stored in the UK, handled to GDPR standards and never used to train models.
Third-party risk
If you are in scope of DORA, we support your ICT third-party risk process: we complete your due-diligence questionnaire and share our SOC 2 report and controls through our Trust Centre.
Decisions you can explain
Every model step and human sign-off is recorded, so you can show how a conclusion was reached, from which documents, and who approved it.
Questions
Common security questions.
Where is our data stored?
In Microsoft Azure’s UK South region. Backups are encrypted and geo-redundant, and the database can be restored to any point in the last 30 days. Beyond that, weekly, monthly and yearly backups are kept for up to two years.
Is our data used to train AI models?
No. Not by us, and not by the model providers. Models are reached only through Microsoft Azure, under enterprise terms that exclude training on customer data.
Which AI models do you use?
Models from OpenAI and Anthropic, chosen for each task and reached only through Microsoft Azure. We never call the model companies’ own services, and every output carries a citation to its source document and page.
Can other clients see our data?
No. Each client’s data is held in its own dedicated database, and every request is tied to a single client before any data is read.
Can we see your SOC 2 report?
Yes. Request it from our Trust Centre at trust.yellowbe.io, run by Yellow Be, the company behind the platform. It also lists the controls we operate, kept current by continuous monitoring.
What happens to our data if we leave?
At the end of your contract we return your data in standard formats and then delete it from the platform.