Trust & security
Your customers' data was meant to stay safe, and stay yours.
Who is accountable, where the data goes, who owns the result and what the rules require.
Plain answers, before you have to ask.
How we work with you
- A senior lead: a complete team from day one, led by a senior AI engineer who stays on your project and is accountable for it.
- Weekly demos: you see working software every week, not status slides.
- Problems raised early: if something is late, unclear or not working, you hear it from us first, with options.
- Fixed-price milestones: the first milestone is fixed-price and defined by the examples it must pass.
- AI-specific contract terms: you own the code, IP and data. We state which model providers are used, that your data is not used to train models, and how prompts, evaluation sets and logs are handed over.
Security and compliance
Every project uses the same controls: least-privilege access for agents and people, audit logs of every action, US or EU hosting, no training on your data, and model providers under matching agreements. See how we build: security and data for the detail.
Healthcare
- HIPAA: we work as a business associate and sign a BAA. There is no such thing as HIPAA certification, so we do not claim one; we show you the safeguards instead.
- BAA chain: PHI only goes to model providers and cloud services that have signed a BAA covering that use. We list every subprocessor.
- PHI handling: minimum necessary data per agent, encryption in transit and at rest, masked data in development and testing, and access logged.
- Clinical safety: agents do not diagnose. Red-flag symptoms are routed to clinicians by rule, and that routing must pass at 100% in evaluation before release.
Europe
- GDPR: we act as processor under a DPA, support data subject requests, and keep records of processing.
- Transfers: where engineers in India access personal data, we use Standard Contractual Clauses and a transfer impact assessment, and keep access to the minimum; many projects use masked data so no personal data leaves the EU.
- EU hosting: systems, data and logs can stay in EU regions, including model endpoints where available.
- EU AI Act: Article 50 transparency duties apply from August 2026: people must be told when they interact with AI, and some AI-generated content must be marked. Obligations for Annex III high-risk systems are expected from December 2027. We assess each system's risk class at the start and design for the obligations that apply.
Financial services
- Audit rights: you and your regulators can audit our work on your systems, as agreed in the contract.
- Explainability: every agent decision records the inputs it saw, the rule that applied and the person who approved it. Decisions that affect customers, such as credit, are made by rules and people, with the model only preparing the case.
- Model risk: evaluation sets, thresholds and results are documented per release so they fit your model risk management process.
Which moment would you fix first?
Tell us where your customers wait today. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with the right team.
We use essential cookies to run this site. With your consent we also use analytics cookies to see which pages help. Privacy policy