Product · How it works

    Structured integrity infrastructure for high-stakes sessions.

    ScreenComply provides system-level signal analysis and audit-ready reporting designed for institutional governance — from the live desktop to the uploaded recording.

    SDVOSB certified · SOC 2 Type 2 examination in progress

    Scope of monitoring

    Exactly what we analyze. Exactly what we never touch.

    ScreenComply monitors system-level signals during virtual sessions — and documents the boundary in every report, so your privacy review is as fast as your security review.

    Signals analyzed

    • Active process and application detection
    • Browser tab and extension monitoring
    • Connected device and peripheral inventory
    • Gaze direction and behavioral patterns
    • AI assistant and overlay tool detection

    What is not accessed

    • No private file content or documents
    • No keystroke logging
    • No personal browsing history
    • No persistent background monitoring
    • No access to non-session applications
    Report output: every session produces a structured integrity report containing risk classification (Low / Moderate / Elevated / High), confidence scoring, timestamped evidence tables, risk modifiers, and scope documentation detailing what was and was not monitored.

    Architecture overview

    From live session to sealed verdict.

    Virtual Session

    ANY VIDEO PLATFORM

    Signal Capture

    SYSTEM-LEVEL ONLY

    Classification Engine

    STRUCTURED MODELS

    Integrity Report

    AUDIT-READY VERDICT

    The reports

    This is what lands on your desk after every session.

    Two deployment modes, two report styles — the Desktop Agent's deterministic restricted-tool verdict, and the Browser API's behavioral analysis with an annotated timeline. Both carry timestamped evidence your committee can act on.

    AGENT REPORT · MACOS · 2M 7S · 8 SNAPSHOTSBinaryScore
    Session tampered. The agent was killed before the session ended; captured data represents only what was recorded up to the termination point.
    11 AI TOOLS RUNNING1 REMOTE-ACCESS TOOLFLOATING OVERLAY · 0% ALPHA

    8 restricted tools found.

    Actual usage state shown. Persistent means the tool ran throughout the session.

    Cluely (New) Helper (Renderer)seen in 35 snapshots · 30s to 2m 6sPERSISTENT
    Wispr Flow Helperseen in 28 snapshots · 30s to 2m 6sPERSISTENT
    CursorUIViewServiceseen in 7 snapshots · tenant rulePERSISTENT
    parsecdremote access · pattern matcherPERSISTENT

    DESKTOP AGENT · BINARY VERDICT REPORT

    SESSION ANALYSIS · UPLOADED RECORDING · 26 FRAMES
    30
    Several signals suggest reviewing the session.NEEDS REVIEW
    SIGNIFICANT OFF-SCREEN BEHAVIORLOW HAND VISIBILITYPOTENTIAL SECONDARY DEVICE USE
    0:00looking away · ~3s0:04.3
    The participant looked away from the screen for a substantial portion of the video. Hand visibility was notably low compared to the cohort, suggesting potential use of secondary devices.

    BROWSER API · BEHAVIORAL ANALYSIS REPORT

    Tamper-evident by design: if the agent is killed mid-session, force-quit, or the machine sleeps, the report says so explicitly — incomplete data is never presented as a clean session.

    Verdict mode

    Score when you want nuance. Binary when you need legal clarity.

    How the report's headline reads is a tenant-level policy you control — and individual events can override it.

    Score (0–100 + bands)

    A 0–100 integrity score with Low / Moderate / Elevated / High bands and per-category confidence. Best for cohort triage and trend analysis across thousands of sessions.

    INTEGRITY SCORE: 30/100 · NEEDS REVIEW

    Binary (Allowed / Restricted)

    The headline renders only "Allowed" or "Restricted tool detected" — a deterministic finding backed by process evidence. Built for integrity hearings and legal proceedings where a score invites argument and a fact doesn't.

    VERDICT: RESTRICTED TOOL DETECTED

    Detection rules

    57 system rules. Always on. Yours to extend.

    The system blocklist updates continuously as new tools ship. Add anything specific to your tenant — or intentionally allow an entry, like ChatGPT for an open-book course. Per-event overrides included.

    CluelyInterviewCoderLockedIn AIFinal Round AIInterview SolverUltraCodeParakeet AIInterviewBuddyVerveAIInterview SidekickAI Interview CopilotLeetCodeMeChatGPTClaudeCopilotGeminiPerplexityCursorWindsurfCodeiumTabnineOtter.aiFirefliesGranolaNotion AIRewind.ai+ 31 more

    Server-side rules update without an agent redeploy — when a new tool launches, every tenant is covered the same day.

    Integrations

    Wired into the calendar, the meeting, and the LMS.

    Events land on your calendar, meeting links are auto-generated, recordings flow into analysis automatically — and candidates never authorize anything.

    G

    Google Calendar + Meet

    ● LIVE  CALENDAR SYNC · AUTO-CREATE MEET LINKS

    • Create an event and a fresh Meet URL is minted — candidates just see a meeting link in their invite.
    • A Meet recording finishes and we pull it from the organizer's Drive straight into the video analysis pipeline.
    • Existing calendar events you didn't create through ScreenComply are never read or modified.
    Z

    Zoom

    ● LIVE  AUTO-CREATE MEETINGS · RECORDING WEBHOOKS

    • Create an event and a scheduled Zoom meeting is minted on your account automatically.
    • A cloud recording finishes, Zoom fires the webhook, we pull the MP4 and run it through analysis — the report lands on the original event.
    • Candidates don't authorize Zoom; existing meetings on your account are never read, edited, or canceled.
    L

    Canvas · Blackboard · Moodle

    ● LIVE  LTI 1.3 · AUTO-LAUNCH FROM ANY QUIZ

    • Instructor adds ScreenComply to a quiz; when a student opens it, the LMS performs an OIDC handshake and the desktop agent auto-launches, already authenticated.
    • No magic-link emails, no separate accounts for students.
    • Each launch maps to course and assignment, so re-running a quiz lands in the same event with the same roster.

    Microsoft 365 (Outlook Calendar + Teams meetings) — coming soon.

    Enterprise deployment

    Governance controls your IT and compliance teams will recognize.

    Works alongside existing platforms

    Operates with the standard video conferencing and assessment tools already in use.

    Structured session signal capture

    System-level signals only. No private file access, no keystroke logging.

    Role-based review access

    Configurable permissions for hiring managers, compliance officers, and auditors.

    Audit log documentation

    Complete audit trail of all system access, report generation, and review actions.

    Secure data handling

    Encrypted in transit and at rest. Configurable retention policies. Session-scoped only.

    Enterprise deployment ready

    Architecture documentation and security review materials available for IT and procurement.

    The difference

    Without structured integrity infrastructure vs. with ScreenComply.

    WITHOUT

    • No audit trail for session integrity
    • No structured session documentation
    • No signal classification or evidence capture
    • Governance and compliance exposure

    WITH SCREENCOMPLY

    • Structured evidence tables with timestamped signals
    • Risk classification scoring (Low → High)
    • Confidence scoring per signal category
    • Audit-ready documentation for governance review
    Aligned with enterprise governance workflows: each report includes scope documentation, classification methodology references, and evidence tables formatted for committee review and regulatory documentation.

    Full-spectrum detection

    Covering the 3 critical points of AI cheating.

    Hidden AI tools, secondary devices, and AI-generated text are how candidates actually cheat today. ScreenComply detects all three.

    01

    Hidden AI Tools

    CLUELY · INTERVIEWCODER · LOCKEDIN AI · OVERLAYS

    The fastest-growing form of cheating. These tools are engineered to be invisible to screen share — transparent always-on-top windows, browser extensions, and background processes that feed answers in real time.

    HOW SCREENCOMPLY DETECTS IT

    • Desktop AI app detection — process-list scanning for Cluely, Cursor, Interview Coder, and named invisible overlays
    • Overlay / always-on-top window detection — catches transparent windows hidden from screen share
    • Browser extension enumeration — blocklist of Copilot, GPT-in-browser, and 120+ AI assistants
    • Remote access tool detection — TeamViewer, AnyDesk, Parsec, RustDesk, and similar
    • Clipboard & paste-event monitoring with keystroke cadence to distinguish typing from paste-and-modify
    02

    Secondary Devices

    PHONES · TABLETS · SECOND MONITORS · OFF-CAMERA COACHING

    Candidates increasingly run ChatGPT on a phone just out of frame, or read answers off a second monitor. ScreenComply catches both the device and the behavior that gives it away.

    HOW SCREENCOMPLY DETECTS IT

    • Mobile phone in-frame classifier with handling-gesture signals
    • Multi-monitor enumeration at the OS level — stronger than gaze inference alone
    • Secondary device network sniffing — detects traffic from devices on the local network
    • Sustained gaze divergence — head pose plus on-screen gaze vector, single glances filtered
    • Unexpected second voice and ambient audio analysis catch off-camera coaching
    • Remote assistance composite flag — reading-while-answering with synchronized audio and gaze
    03

    AI-Generated Text

    CHATGPT ANSWERS · PASTED RESPONSES · SYNTHETIC WRITING

    Whether pasted into chat, dropped into a coding environment, or read aloud, AI-generated text leaves a statistical fingerprint. ScreenComply analyzes the artifact and the behavior around its delivery.

    HOW SCREENCOMPLY DETECTS IT

    • AI-generated document detection — perplexity, burstiness, and semantic-drift classifiers calibrated against RAID
    • AI-generated code detection — stylistic consistency, naming entropy, and comment cadence
    • Paste-event logging distinct from typing, with optional content capture by policy
    • Keystroke dynamics — inter-key timing separates natural typing from paste-and-modify
    • Reading-while-answering detection — audio and gaze synchronized to a hidden text source
    Plus: Generative AI fraud — deepfake faces and cloned voicesLess common today than hidden-tool cheating, but rising. ScreenComply also runs frame-level deepfake classifiers, 3D liveness with rPPG, and vocoderprint analysis against RAID benchmarks to catch synthetic media when it appears.

    Your data, your rules

    Fully flexible data retention and privacy controls.

    ScreenComply adapts to your organization's data governance requirements, from zero retention to full archival.

    From read-only to full retention

    Need zero data retention — not even reports? ScreenComply can operate in read-only mode where all signals are analyzed in real time and discarded immediately after the session. No data touches disk. Or retain everything — session data, integrity reports, behavioral analytics — and reuse it for benchmarking and continuous improvement of your integrity program.

    Evidence-backed assessments

    Every classification — Suspicious, Highly Suspicious, or No Issues Detected — is backed by a detailed evidence report, built on statistical analysis, rules-based determinations, and computational anomaly detection. Text and audio models are benchmarked against the hardest widely available datasets, including RAID.

    Customer-first integration

    Every deployment is configured to your privacy requirements, data residency needs, and integration architecture. We work with your legal, IT, and compliance teams to fit ScreenComply into your existing governance framework — not the other way around.

    Privacy by design

    No keystroke logging. No screen recording. No personal file access. ScreenComply collects only the system-level signals necessary for integrity classification — nothing more. Encryption at rest and in transit on SOC 2 aligned infrastructure.

    Get started

    See the platform in action.

    Walk through a live detection, a real integrity report, and the governance controls with our team.