Skip to content
ReleasePoint by Data Machines
Data Machines Sign in

Human-led · Machine-assisted

Find what matters.
Defend what you withhold.
Release with proof.

ReleasePoint turns mixed-format records into a controlled review workflow — automating the search for sensitive content while keeping every legal decision with a named Government reviewer.

See the workflow Where AI actually helps
Mixed-format intake Human decisions Verified output
FOIA 2026-0455Meridian lease records
ANALYSIS COMPLETE
Analyze
2Review
3Approve
4Release
14documents
74suggestions
0auto-withheld
21.0sbatch wall clock
SUGGESTED REDACTIONRULE

Personal contact · (b)(6)

r.calloway@meridianproperties.com
94%deterministic match

WHY IT WAS FLAGGED

Non-agency email address; possible personal contact information.

Human judgment requiredReleasePoint found evidence — not a withholding decision.
Analyze the whole requestPDF, Office, email, images and containers
No automatic withholdingEvery proposal requires a human decision
Two-person releaseSubmitter and approver are always distinct
Proof before downloadBurned redactions and forensic verification

The operating problem

Review is where FOIA work slows down.

ReleasePoint attacks the middle of the process: finding candidate content, keeping decisions consistent across a request, documenting the legal basis, and proving the released file contains what the reviewer intended.

01

Read at machine speed

Normalize the request, OCR scans, and surface candidate redactions across every supported format.

02

Decide at human depth

Put the evidence, exemption, and required legal determination in front of the reviewer together.

03

Release with evidence

Bind approval to the decision set, burn the output, verify it, and package the audit trail.

One accountable pipeline

From raw records to defensible release.

Choose a stage to see what the system does — and what it deliberately leaves to people.

01 · Analyze

One request in. A review plan out.

ReleasePoint normalizes mixed records, OCRs scans, and runs the methods authorized for the environment. Every candidate is labeled by how it was found.

    Request analysisCOMPLETE
    14records normalized
    13ready to review
    1needs attention

    Pattern rulesIdentifiers, contact data, financial terms

    58

    VisionHandwriting and signature regions

    16

    Semantic reviewDisabled for this environment

    OFF

    Human additionsManual findings remain distinct

    3
    21.0 seconds observed from upload start to review-ready suggestions.

    Assistance you can interrogate

    Automation does the reading.
    People make the call.

    ReleasePoint does not hide every method behind the word “AI.” The reviewer sees whether a finding came from a deterministic rule, computer vision, optional semantic analysis, or a person.

    Your boundary sets the rules.Semantic review remains off unless an authorized endpoint is configured. When off, document content does not leave the system.

    RULES

    Repeatable pattern detection

    Identifiers, financial terms, agency-configured policy logic, and exact-match validators.

    DETERMINISTIC
    VISION

    Visual region detection

    Signatures, handwriting, and content that text extraction cannot reliably reach.

    MODEL
    SEMANTIC

    Contextual review

    Optional analysis for deliberative passages and contextual omissions at an approved endpoint.

    OPTIONAL
    HUMAN

    Legal decision and rationale

    The final accept, reject, modification, exemption basis, and segregability determination.

    AUTHORITATIVE

    Defensibility by construction

    The release package proves what happened.

    A blacked-out PDF is only the visible result. ReleasePoint preserves the decisions and verification needed to explain that result later.

    Approval is bound to the exact redaction set. Change the set and the approval no longer matches.

    release_package.zipGenerated after verificationREADY
    Released documentsBurned output — not overlays
    Exemption logRegion, code and decision basis
    Verification reportsLeak checks and output hashes
    Audit extractWho did what, and when
    Forensic verification passed before package generation

    Government control model

    Controls that show up in the workflow.

    Security and governance are not a footer claim. ReleasePoint reports the configured posture inside the product and enforces critical gates below the interface.

    Agency boundary

    Run locally with external semantic services disabled, or restrict calls to an explicitly approved endpoint.

    Separation of duties

    The person submitting a decision set cannot approve the same release.

    Tamper-evident output

    Download and rendering recheck the released file against its verification report.

    Enforced retention

    Agency-configured retention begins at closure, with authorized holds and audited sweeps.

    Accessible review

    Keyboard-operable review and baseline accessible output support; formal PDF/UA scope remains deployment-specific.

    Observed performance

    Wall-clock batch timing is reported separately from per-document and detection-only measurements.

    ReleasePoint reports the environment it is running in. Accreditation, production authorization, and agency policy configuration remain deployment responsibilities — not marketing claims.

    See the entire path

    Bring a mixed-format request.
    We’ll show you the release proof.

    Evaluate ReleasePoint on the work that matters: intake, assisted review, human legal decisions, four-eyes approval, burned output, and verified release.

    Contact Data Machines

    [CONTACT EMAIL] · [PHONE] · UEI [UEI] · CAGE [CAGE CODE]