Acremind is currently offered as an early-access pilot for commercial real estate brokers and managers. This policy describes the data we collect, where it is stored, how long it is retained, and the controls you have over it during the pilot. For a longer, plain-language walkthrough see the "Data Privacy, Retention, and Residency" guide in the in-app Help library.

• Account and profile: email, display name, persona (CRE Manager), organization, and Firebase Authentication identifiers. • Workflow content you create: Spaces, deals, listings, tasks, LOIs, DCF runs, lease abstractions, tour packages, brand settings, Studio artifacts, shareable links, and anything you attach. • Agent conversations: text turns, voice transcripts, and images or audio you send to the multimodal AI assistant. • Operational telemetry: authenticated request logs, agent action audit logs, evaluation scores, and crash/error traces. We use this for reliability, security, and support only.

We use your data to operate the product you explicitly invoke (searches, agent runs, artifact generation, shareable links), to secure and audit the platform, and to improve reliability and safety. We do not sell pilot data, we do not use it to train foundation models, and we do not share it with advertisers.

All pilot data is stored in Google Cloud Platform, United States regions, with us-central1 as the primary region for Cloud Run, Cloud SQL, Firestore, Redis, BigQuery, and Cloud Storage. AI requests run on Google Vertex AI (us-central1) and, for specific preview models, the global Gemini endpoint. These are governed by Google Cloud's enterprise data-processing terms.

Retention is enforced automatically by storage-level lifecycle policies, not by the honor system: • Agent audio, image uploads, evaluation artifacts, and client feedback: 90 days. • User-uploaded documents (deals, leases, supporting files): 2 years. • Generated artifacts (LOIs, reports, PDFs): 1 year active, then archived and deleted at 3 years. • Operational HTTP request logs: 365 days. • Standard agent action audit trail: 2 years active, then archived to Parquet in Cloud Storage. • Financial audit trail (LOIs, DCFs, draft/publish listing): 7 years, per financial-records norms. • Forensic audit-log copy: 7 years in a Cloud Storage Object-Lock bucket — immutable, cannot be altered or deleted even by administrators. • Agent session state in Redis: ephemeral, cleared on deletion or session end. • Profile, deals, and listings in Firestore: retained until you request deletion.

• All traffic is TLS. Authenticated requests carry a Firebase JWT; REST calls additionally require an API-gateway key. • Agent prompts and responses are filtered by Google Cloud Model Armor with a Cloud DLP de-identification template. High-sensitivity tokens (SSN, PAN) are permanently masked; contextual PII (names, phone numbers) is reversibly tokenized with Format-Preserving Encryption backed by Cloud KMS. • An LLM-as-Judge plugin screens inbound prompts for prompt injection before tool-calling code runs. • A PII-sanitizing filter scrubs SSNs, emails, phone numbers, JWTs, and credit-card patterns from structured logs before they leave the process. • Every privileged action is written twice: once to Cloud SQL, and once to an Object-Lock bucket for forensic immutability.

Pilot data is processed by Google Cloud (infrastructure, Vertex AI, Gemini, Firebase, BigQuery) and Firebase Authentication as sub-processors under Google's enterprise data-processing terms. We do not share pilot content with any other third party. If your organization requires a signed data-processing agreement before expanding usage, contact pilot support.

You can request access to, export of, or deletion of the workflow content tied to your pilot account at any time. Deletion is executed by our Right-to-be-Forgotten (RTBF) pipeline across Firestore, Redis, and Cloud Storage, and deletion fulfillment is tracked in an auditable compliance view. Two items survive a deletion request, by design: the financial and forensic audit trail (required for fraud investigation and regulatory evidence) and aggregated, de-identified telemetry that no longer identifies you. If that tradeoff matters to your use case, tell us before loading sensitive material.

Acremind's safeguards are designed for CRE workflows. During the pilot, please do not upload regulated PHI, children's data, payment-card primary account numbers, or government-issued identifiers, and do not upload any counterparty material you are contractually barred from sharing with a cloud vendor.

We will update this policy as the product moves from early access to general availability. Material changes will be announced through the pilot support channel before they take effect.

Privacy or data-handling questions: privacy@automatixai.live. Pilot support for access, export, or deletion requests: use the pilot support channel shared in your rollout email.