How TrialScope collects, uses, stores, and protects your data — written in plain language because we think clinical research professionals deserve clarity, not legalese.
TrialScope collects three categories of data:
Account information. Your email address, password (stored as a one-way cryptographic hash, never in plaintext), professional role (if provided), and basic account metadata. You may also sign in via LinkedIn OAuth, in which case we receive your basic profile information from LinkedIn in lieu of storing a password.
Audit content. Documents you upload (PDFs, DOCX, etc.), text you paste, images you capture, trial metadata you enter, and the AI-generated audits produced from this content. This is the "audit substrate" — the material we process to produce your audits and store so you can return to them.
Calibration contributions. Corrections you make to audit findings, expert notes you provide, ratings, and the consent scope you select for each contribution. This is how TrialScope learns from expert users to produce better audits over time.
We use light analytics (PostHog) to understand which features are used and where users encounter friction. This collects standard web telemetry — page views, click events, browser type — and is associated with your account ID, not any external identifier.
To produce your audits. Documents and metadata you provide are sent to Anthropic's Claude API for processing. Anthropic's API terms prohibit them from using API-submitted content to train their models. Audits return to TrialScope, are saved to your account, and become accessible only to you.
To improve the platform. Calibration contributions you make — corrections, notes, ratings — are used to improve audit quality for all users, scoped to the consent level you select per contribution. You always control the scope.
What we do NOT do: We do not sell your data. We do not share your documents with third parties beyond Anthropic for audit processing. We do not use your documents to train AI models — neither ours nor anyone else's. We do not surveil your work for our own purposes.
Our consent framework is called the Strategic Middle Path: you own your content, you control the scope of each calibration contribution, and improvements you contribute benefit the platform's collective expertise without compromising your data ownership.
You. All your documents, audits, corrections, and account data are accessible only to you. Row-level security policies in our database enforce this at the storage layer, not just the application layer — meaning even if our application code had a bug, the database itself would prevent cross-user data access.
Anthropic. When you submit an audit, your documents and metadata are sent to Anthropic's Claude API for processing. Anthropic's API processes the request and returns the audit. Anthropic does not retain submitted content for training and has its own data handling practices governed by their API Terms of Service.
Supabase. Our database and authentication provider. Your data is stored in Supabase's infrastructure (US-region). Supabase has its own privacy commitments and is SOC 2 Type II certified.
Vercel. Our hosting provider. Web traffic flows through Vercel's infrastructure. Document content is not stored by Vercel — it passes through during the API request lifecycle.
TrialScope administrators. The TrialScope founder may access aggregate platform metrics, calibration libraries (de-identified), and individual records only as required to investigate support requests, security incidents, or compliance obligations. Administrative access is logged.
Account data: Retained for as long as your account exists. Deleted when you delete your account.
Audit content: Retained for as long as you keep the audit. You can delete individual audits at any time from your dashboard, which removes the document content, the audit output, and associated metadata.
Calibration contributions: Calibration data you've contributed under the Strategic Middle Path framework remains in the platform's calibration library to benefit future audits, but is de-identified from your account if you delete your account. Contributions you marked as confidential are subject to additional handling per the consent scope you selected.
Backups: Database backups are retained for 30 days for disaster recovery. Deleted data may persist in backups for this period before being purged.
You have the right to:
If you are an EU resident, the General Data Protection Regulation (GDPR) provides additional specific rights. We will honor them. Contact us via the email above to exercise these rights.
TrialScope is not a HIPAA-covered entity and is not designed to handle Protected Health Information (PHI).
Please do not upload documents containing patient identifiers, individual subject-level data, identifiable medical records, or any content protected by HIPAA. TrialScope is designed for trial methodology review using protocols, statistical analysis plans, investigator brochures, and similar planning documents — not patient data.
If you upload PHI in error, contact us immediately at robert@trialscope.ai and we will work with you to delete the affected content from our systems and assess any further mitigation needed.
All data transmitted to and from TrialScope is encrypted in transit using TLS. Stored data is encrypted at rest in Supabase's infrastructure. Passwords are stored only as one-way cryptographic hashes (bcrypt) — never in plaintext, even in our database. We support password sign-in and LinkedIn OAuth; password reset uses a time-limited tokenized email link.
We have not yet completed formal security certifications (SOC 2, ISO 27001, etc.). These are on our roadmap. In the interim, our infrastructure providers (Supabase, Vercel, Anthropic) hold their own certifications which apply to the portions of our stack they operate.
We may update this policy as the platform evolves. Material changes will be communicated to registered users by email at least 14 days before they take effect. The "Last updated" date at the top of this page always reflects the most recent revision. Historical versions are available on request.
Questions about this policy, requests to exercise your rights, security concerns, or anything else related to data handling: robert@trialscope.ai.
We aim to respond to substantive requests within 5 business days during early access. As the platform scales, we will publish a formal SLA for data-handling requests.