A Verified Account Doesn’t Prove a Human Is Present
Research Initiative · 2026
Identity checks can authenticate an account without showing that a person is actively participating. HUMNLABS explores browser-local interaction signals as descriptive evidence—without requiring identity or claiming proof of humanity.
About 3 minutes · No identity required · Experiment samples are not transmitted
Verifying identity is not the same as verifying humanity.
But authentication alone does not establish that a human is actively present in an interaction.
HUMNLABS explores whether interaction-derived signals can contribute to structured evidence coverage without requiring identity.
The challenge is no longer building intelligence.
The challenge is building trust when certainty is impossible.
As AI systems become more capable, the internet increasingly struggles to distinguish between humans, AI agents, automated systems, and synthetic identities. Most existing systems ask: "Who are you?" HUMNLABS asks whether a session contains sufficient browser-local interaction evidence for an educational coverage summary. The result is deterministic and non-numeric; it does not establish humanity, identity, or a probability.
Why This Matters Now
The digital environment is crossing an irreversible inflection point. Four forces drive our investigation into privacy-preserving trust signals.
AI Agent Proliferation
Autonomous systems scale exponentially. HUMNLABS evaluates how AI agents may drive the majority of web traffic, acting, executing, and transacting without human oversight.
Deepfakes & Synthetic Media
Advanced generative models synthesize high-fidelity voice, video, and text. Our research investigates how this renders traditional methods of manual check-based verification increasingly obsolete.
Coordinated Bot Networks
Automated systems mimic organic behaviors at scale. We explore how coordinated networks flood communication channels, prompting the need for privacy-preserving presence indicators.
Synthetic Identity Collapse
Completely fabricated digital personas can be operated autonomously. We investigate how synthetic identity may challenge trust in online relationships, platforms, and digital systems.
How can browser-local interaction evidence be summarized without requiring identity?
What We Are Researching
HUMNLABS focuses on the intersection of AI, privacy, and digital trust where certainty may be impossible.
Session Evidence Coverage
Research into whether browser-local interaction signals can produce structured, descriptive session summaries.
Trust Signal Infrastructure
Investigating privacy-preserving signals that may support confidence-based digital interactions.
AI Agents & Human Presence
Research into how systems may distinguish interaction characteristics without requiring identity.
Human Signature Research
Exploring whether combinations of motor, behavioral, semantic, and contextual signals may form privacy-preserving interaction signatures.
Infrastructure for Digital Trust
HUMNLABS explores infrastructure that may help digital systems reason about human presence without requiring unnecessary identity or personal data.
-
Session Evidence Coverage
Investigating how browser-local interaction signals can contribute to an educational evidence-coverage summary without establishing identity, humanity, or trustworthiness.
-
Privacy-Preserving Signals
Exploring cryptographic and zero-knowledge methods to pass signals without sharing personal data.
-
Evidence-Coverage Outputs
Prototyping deterministic evidence-coverage models that make limits and missing evidence explicit rather than issuing binary classifications.
-
Explainable Signal Reasoning
Exploring transparent explanations of which evidence signals were observed, limited, or insufficient.
-
Protocol-Level Research
Compiling architectural specifications for a long-term vision of a decentralized trust protocol.
How We Evaluate Evidence
Our research separates available evidence from unsupported conclusions. The public demonstration summarizes browser-local session observations, while the isolated laboratory tests deterministic evidence rules using synthetic fixtures.
Collect Session Samples
During the public Experiment, reaction, movement, and typing samples exist temporarily in browser memory. The interaction samples are not stored, transmitted, or linked to an identity.
Validate Sample Structure
Deterministic checks confirm whether a sample is usable, incomplete, excluded, or invalid before any summary is produced. These checks evaluate data structure, not whether someone is human.
Derive Descriptive Observations
Usable samples can produce limited session observations such as median response time, collection duration, or mean input interval. These measurements are descriptive and are not identity or humanity assessments.
Report Evidence Coverage
Coverage records how many task summaries were available. Missing input or an accessible alternative changes coverage only because no observation was produced; it is not treated as suspicious behavior or a penalty.
Abstain from Unsupported Scores
HUMNLABS does not convert uncalibrated observations into a humanity percentage or a human-versus-bot verdict. Evidence limits remain visible instead of being hidden behind an authoritative score.
Separate Demo from Laboratory
The public Experiment is an educational browser demonstration. The private Trust Engine laboratory is separate, deterministic, synthetic-only, uncalibrated, and not integrated with production.
Current Research Boundary
HUMNLABS has not performed real-participant calibration or established accuracy, false-positive, or false-negative rates. Future participant research would require separate governance, privacy assessment, accessibility review, and explicit approval.
Current Research Reality
HUMNLABS distinguishes working laboratory code from future research goals. This is the current status of our deterministic Trust Engine research.
Deterministic Lab Baseline
We built an isolated research engine that computes derived movement, typing and reaction observations. Repeated analyses produced byte-identical output in the recorded laboratory environment. No AI model or LLM participates in the evaluation path.
Synthetic-Only Evaluation
The engine was evaluated across 213 deterministic synthetic scenarios covering sampling frequency, timer quantization and sample loss. No real participant data was collected or processed.
Coverage, Not Classification
The engine reports evidence availability rather than a binary human-or-bot label. Missing inputs and accessibility exclusions reduce coverage without creating anomaly penalties or artificial certainty.
Calibration Not Performed
The laboratory engine does not output a calibrated trust score. Real-world behavioral distributions remain unmeasured, and its provisional risk thresholds remain unvalidated.
Demonstration vs. Laboratory Engine
The public interactive Experiment is an educational demonstration of browser-local interaction signals. It outputs a non-numeric educational signal summary, is not produced by the private Trust Engine laboratory, and is not proof of humanity, identity verification or calibrated confidence. The laboratory remains separate from the production website.
Active Research Initiatives
HUMNLABS is currently developing the following research publications and experiments. These are not yet available for download.
Human Presence & Trust Report 2026
A research report exploring human presence, confidence, identity, AI agents, and digital trust.
Human Presence Experiment v0.1
An experimental demonstration exploring whether interaction signals can contribute to confidence that a human may be present.
Human Trust Layer Framework
A conceptual framework for privacy-preserving trust signals and confidence-based digital interactions.
HUMNLABS Principles
A research foundation defining the limits, responsibilities, and privacy principles of human presence estimation.
What HUMNLABS Does Not Claim
Privacy-preserving interaction research is an exploratory inquiry, not proof of humanity.
Explore the Human Presence Experiment
Can interaction-derived signals contribute to confidence in human presence without requiring identity, biometrics, or personal data?
Demonstration Notice
The public interactive Experiment is an educational demonstration. Its illustrative result shows how selected timing signals can be presented; it is not proof of humanity, identity verification, calibrated confidence, or output from the deterministic Trust Engine v0.1 laboratory.
Access the Human Presence & Trust Report 2026
Explore HUMNLABS research into human presence, privacy-preserving trust signals, confidence in human presence as a research question, and the limits of identity in an AI-mediated internet.
Independent research publication. No product. No hype.
By registering, you consent to Smyrna Digitale using your email solely to notify you when the report is released. Your submission is processed through Formspree. You can withdraw and request deletion at any time. Read the Privacy Policy.