No training on your data
We don't train AI models on your data, and neither do our AI providers.
Your data and Barry
What Barry sees, and where it goes
Barry and BlackHyve's other AI features help answer questions, draft paperwork and organize project information. Here is what each feature uses, where it goes and what happens afterward.
Download the PDF- Your teamStart a workflow
- BlackHyvePrepare the context
- AI provider or AWS modelsProcess the request
- BlackHyveReturn the result
- Your teamReview or use it
Some features start automatically. Retention depends on the processing path below.
The short version
Barry sees what you see
Barry only reads company records you have permission to access.
Requests aren’t stored*
Most AI requests use Zero Data Retention at the model provider.
Chat can’t change your data
Barry can look things up in chat, but can’t edit your schedule, budget or change orders.
* File, batch and video processing have separate retention rules. Provider terms and diagnostic retention are explained below.
What each feature reads
| Feature & workflows | What it uses | AI processing |
|---|---|---|
Ask questions; request tasks | Messages, conversation history and project records returned by tools: schedules, captures, incidents, boards, change orders and more. | AI providerText / media |
Contract schedulesRetention exception Import, organize, compare, report & mark Exceptions: PDF import & revision matching: Anthropic code-execution data up to 30 days. | Schedule PDFs, tasks, dates, trade/location context and your instructions. Matching and classification use schedule task context. | AI providerFiles / text |
Summarize; clean up answers | Day's activities, progress, blockers, log questions and answers. | AI providerText |
Create activities; match codes & crew | Spoken-plan transcript, dates, cost codes and candidate crew names. Speech recognition is a separate step. | AI providerText |
Field captures & voiceRetention exception Dictate, transcribe, translate, describe & scope Exceptions: Video transcription, translation & scope reprocessing: files up to 48 hours; abuse-monitoring logs up to 55 days. | Microphone audio for dictation; complete videos for AI processing, plus event/project context; transcript text for short descriptions. Dictation is transcribed on the device; its audio stays there. | Device / AI providersMixed |
Blur faces; mask people; reconstruct | 360 recording frames, plan imagery and recording context. Vision models run inside BlackHyve's AWS job. Originals remain separate. | BlackHyve / AWSLocal models |
Triage, find, draft & discuss | Capture transcripts, descriptions, locations, classifications, linked evidence and conversation history. | AI providerText |
Build, summarize, edit & rewrite | Relevant project records, notes, selected text, report blocks and any attached or linked documents/images. | AI providerText / media |
Draft cards; discuss work | Your instructions, board conversation, cards, statuses, priorities and project context. | AI providerText |
Run flows; draft prompts; morning brief | Flow instructions, project/activity context and records retrieved for the message or brief. | AI providerText / media |
Write narratives; extract items | Linked incidents, captures, work details and supplied pricing documents or images. | AI providerText / media |
Draft an RFI | The question, supplied context and any attached PDF. | AI providerText / media |
Draft notices; explain blockers | Linked incident or stuck-work details, responsible people, companies and project context. | AI providerText |
Extract a paper ticket | The full uploaded PDF, including visible names, signatures, prices and work details. | AI providerDocument |
Import a budget matrix | The uploaded PDF or CSV, financial details, locations, systems and selected import context. | AI providerDocument / text |
Extract, draft, add keywords, stages & systems | Scope text, existing codes and names, descriptions, hierarchy and planning context. | AI providerText |
Read page text & title blocks | Extracted page text and selected image regions from drawings. | AI providerText / images |
Extract from text or drawings | Schedule/task text, project locations, or a supplied drawing/document. | AI providerText / media |
Work & schedule summariesRetention exception Summarize activities, job walks & schedules Exceptions: Schedule batch summaries: 29 days. | Activity and schedule records, dates, progress, locations, grades, notes, captures and blocker details. | AI providerText / batch |
Who processes it, and what they keep
Zero Data Retention means prompts and responses are not stored after an eligible request. It does not delete BlackHyve's saved records.
| Processor | Data & retention | Model training |
|---|---|---|
| Anthropic Claude | Eligible text and directly submitted documents/images use Zero Data Retention. Provider files are deleted after successful or failed processing, with retries. Code-execution data can remain up to 30 days; batch data, 29 days. Legal and malicious-use exceptions apply. | Prompts and responses are not used to train models. |
| Google Gemini | Receives complete field videos. Uploaded files expire after 48 hours unless deleted sooner; separate abuse-monitoring logs can remain 55 days. This route is outside Zero Data Retention. | Our paid Gemini service does not use prompts or responses to improve Google products. Abuse-monitoring data can be used for policy-enforcement models. |
| BlackHyve / AWS | 360 vision models run inside BlackHyve’s AWS environment. Recordings and generated results follow the workspace retention and deletion rules below. | Runs existing models to process data; does not train them. |
| TypeSafe Jev | Provides additional contract-schedule matching and classification, including suggested matches, using task context and candidates. Zero Data Retention is enabled for BlackHyve’s TypeSafe account. | Customer inputs are not used to train or fine-tune models. |
| On-device dictation iOS / Android | Transcribes microphone audio on the device. Dictation audio is not sent to an external speech provider. | On-device inference; no external model training. |
Who can see your data
People on your jobs
Barry works with the same access as the person using it. Someone with access to two jobs gets answers from those two jobs only. Saved chats follow your workspace’s retention and deletion rules.
Never other companies
Other BlackHyve customers do not get access to your company’s records through Barry. Answers stay within your company and the records your permissions allow.
Our AI providers
The AI provider reads the question and relevant job records sent by the feature so it can return a result. The table above lists each provider and its retention and review terms.
PostHog
Records prompts, responses and tool context from instrumented AI workflows for diagnostics. This content is not anonymized and has a 30-day AI-trace retention period. Usage and cost metadata are retained separately.
Diagnostics: Provider Zero Data Retention does not apply to BlackHyve’s PostHog logs. PostHog also has a separate 90-day session-replay retention setting; the 30-day AI-trace period does not govern replays or analytics metadata.
Saved workspace data: Uploads, saved chats, 360 recordings and generated results have no automatic expiry while your workspace is active. Cancelled workspaces are disabled within 30 days, or as your contract specifies; backups can remain for up to one year after cancellation. Request deletion at privacy@blackhyve.com. Earlier deletion is accommodated within technical capabilities, subject to legal and contractual requirements.
Approved evaluations: Specific de-identified data is used only with customer and BlackHyve approval. It is stored on AWS indefinitely; you can withdraw permission at any time to stop use and delete our evaluation copies.
This guide describes data processing and does not amend your agreement or BlackHyve’s privacy policy.
BlackHyve / 18 October 2026
Back to top ↑