No digital system is 100% safe, we take reasonable efforts across multiple dimensions to keep your information as safe as we can. All information is stored in a fully encrypted database, accessible only to the users authorised to access it. Our AI systems are hosted in private instances, away from the companies developing the models, so you can rest assured the data won’t be used for training without your explicit permission.

(Data is AES‑256‑encrypted, stored on secure ISO 27001 certified EU-hosted Microsoft Azure servers, and never used to train external models. Access is limited by permissions and SSO).

We also undertake frequent vulnerability checks across all our vital systems with a mixture of periodic and automatic monitoring.

vBrief's security posture covers infrastructure, operations, data handling, and insurance. Key measures include: EU-hosted Azure (Sweden) with AES-256 encryption and TLS 1.3; OIDC SSO with MFA; row-level permissions; weekly automated vulnerability scanning; Microsoft Defender across infrastructure; a tested incident response plan with GDPR notification timelines; and DPIA screening completed with a designated privacy lead.

No customer data is used for AI training — this is a contractual provision, not just a policy. We are working towards Cyber Essentials Plus and ISO 27001 certification. For a full security overview or interim procurement documentation, contact security@vbrief.ai.

All our systems run on EU servers; and data never trains external models. On-premises models are not currently available, though if this, or bring your own key (BYOK), is a major decision factor for your business, please speak to vBrief and we can discuss how we can accomodate this.

vBrief’s AI is contained to key areas of the workflow, accelerating the user but not making decisions on their behalf. Users verify information before it is converted into output by the AI making it much more traceable. Crucially, it also ensures users can easily review to spot if our AI has missed out any key information, one of the largest potential PI insurance risks when using end-to-end AI applications (commonly known as the omission risk).

Each decision is stored using a tamper‑proof log as a continuous Golden-Thread, helping towards typical PI‑insurer requirements for traceability and auditability.

Yes - AI will show you all its understanding in the original context. The AI instance is completely detached from the systems used to commit user-review decisions.

Complex design decisions require professional judgement. Most AI agents who complete nuanced tasks for users are still very immature, and have a very high error rate due to hallucinations or major misunderstandings. If the AI makes a mistake, the professional using the system is ultimately liable for any harm that comes from it.

At vBrief, we understand your project communications are mission critical and your professional liability is not to be taken lightly. Omission risks are difficult to spot and an unlucky event could leave your insurance in question. Our software uses AI where major efficiency gains are possible, while ensuring major checks and decisions lie with the professionals where the legal liability lies.