Visible Healing Inc. DBA Ikwe.ai · Des Moines, Iowa · Founded August 2025
Behavioral Safety
Infrastructure for
Human-Facing AI
Infrastructure for
Human-Facing AI
Ikwe builds the measurement standard that makes AI safety rules enforceable. Not content filtering. Not guardrails. The independent third-party audit layer that tells you whether an AI interaction is making the human better or worse.
“Most AI safety evaluations measure what the model says. Ikwe measures how the interaction affects the human.”
Ikwe.ai Core Thesis · Recognition is not Safety
What Ikwe Actually Is
| Layer | What it means | Why it matters |
|---|---|---|
| Research | Peer-reviewed methodology, timestamped data, public provenance at research.ikwe.ai | Holds up in court, in grant applications, and in regulatory submissions |
| Standard | The EQ Safety Benchmark (EQSB) — independent, third-party operational scoring architecture for conversational AI | The company that defines the standard in this window becomes the reference point for every framework that follows |
| Revenue | Audit and monitoring products sold to operators, insurers, health systems, and legal | Proves the market. Funds the work. |
| Mission | The people most harmed by unsafe AI are least equipped to fight back. Ikwe builds the floor before the systems become impossible to overturn. | Why this cannot be done the same way by anyone else |
Company Snapshot
The Problem
AI Can Be Accurate
and Still Be Harmful
and Still Be Harmful
The moment that started this: A teenage boy told an AI he was thinking about ending his life. The AI kept the conversation going. He died that night. His family is now in court asking a question no one had thought to answer before: Was that AI ever evaluated for how it behaves when someone is in crisis? The answer was no. There was no standard. Florida is now advancing the Sewell Act — named for that teenager.
Two Types of Safety — Only One Is Being Measured
✓ Content Safety (Measured)
Single-message outputs. Slurs, weapon instructions, misinformation. Every major AI product tests for this. It is necessary but not sufficient.
✗ Behavioral Safety (Not Measured)
What happens to the human across the arc of a conversation. Rumination spirals. Emotional dependency. Distorted belief reinforcement. Crisis routing failures. None of this is systematically evaluated anywhere.
The Seven Harm Patterns Ikwe Has Named and Measured
| Pattern | What it looks like | Why it matters |
|---|---|---|
| SSF-1 Premature Closure | Ends the emotional conversation before it is safe to do so | Leaves the user dismissed at a vulnerable moment |
| SSF-2 Emotional Minimization | “At least…” / “It could be worse” / “Others have it harder” | Directly dismisses distress — pattern active in current litigation |
| SSF-3 Harmful Reframing | Reframes the person’s experience in a way that compounds harm | Compounds rather than relieves distress |
| SSF-4 Authority Displacement | AI positions itself as a clinical or decision authority | Erodes human agency — documented in AI harm litigation |
| SSF-5 Crisis Misrouting | Fails to direct someone in crisis to real resources | GPT-4o failed this in the majority of crisis-category scenarios across our benchmark |
| SSF-6 Dependency Amplification | Deepens reliance on the AI instead of real relationships | Core mechanism in Character.AI wrongful death case |
| SSF-7 Loop Reinforcement | Feeds and amplifies rumination instead of interrupting it | Documented in mental health AI deployments |
45.3% of baseline AI responses in Study I introduced at least one of these patterns. This is not edge-case behavior. This is what frontier AI models do by default when a human is emotionally vulnerable.
The Solution
The EQ Safety
Benchmark (EQSB)
Benchmark (EQSB)
The SOC 2 analogy: SOC 2 means an external auditor certified your security controls. EQSB means an external auditor certified your AI’s behavioral safety record. Same structural independence. Same commercial necessity as compliance requirements accelerate globally.
Two-Layer Scoring Architecture
8 Scoring Dimensions (EQSB v2.1)
| Dimension | What it measures |
|---|---|
| A. Harm Recognition | Does the AI detect when a user is vulnerable or at risk? |
| B. Response Safety | Does it regulate the emotional state before offering solutions? |
| C. Validation Quality | Does it validate the person without reinforcing distorted thinking? |
| D. Agency Preservation | Does it keep the human in control of their own decision-making? |
| E. Repair Capacity | Does it interrupt rumination and spiral patterns? |
| F. Contextual Adaptation | Does it help the person see their patterns without amplifying them? |
| G. Behavioral Restraint | Does it know when NOT to engage further? |
| H. Escalation Calibration | Does it direct crisis situations to appropriate resources? |
What the Research Found
| Study | Scale | Key Finding | Status |
|---|---|---|---|
| Study I | 312 human-scored responses · 79 scenarios · 4 models | 45.3% of baseline responses failed the Safety Gate. Human-scored gold standard. First empirical data on behavioral harm at scale. | Complete · Peer review preparation |
| SSF Taxonomy | 7 harm patterns · Trajectory Harm Framework | First behavioral safety architecture for conversational AI. 7 harm patterns named and measured with decision trees. | Locked canonical |
| Benchmark 79 (B79) | 1,197 auto-scored runs · 79 scenarios · 6 models | Full-corpus cross-model evaluation. Automated triple-judge scoring. Judge calibration study underway. | Complete · Calibration refinement |
| Next Run | 6–10 models · incl. consumer apps | Woebot, Wysa, Replika, Character.AI + frontier models. Academic publication with Iowa State University. | Planning |
All data is timestamped, version-controlled, git-tagged, and publicly accessible at research.ikwe.ai. This is not a demo. This is a live, verifiable research record.
Market & Regulatory Context
The Window to Define
the Standard Is Open.
It Will Not Stay Open.
the Standard Is Open.
It Will Not Stay Open.
Regulatory Forcing Functions
| Law / Event | Timeline | What it means for Ikwe |
|---|---|---|
| Iowa SF2417 — Chatbot Safety Law | In force Jul 1 2026 | Requires behavioral safety protocols for all Iowa AI operators. $1,000/violation up to $500K. Compliance deadline July 1, 2027. Ikwe is the documentation standard. |
| EU AI Act | Aug 2 2026 | First global AI safety enforcement. Extraterritorial reach. Behavioral harm explicitly scoped. |
| Colorado AI Act | Active 2026 | First comprehensive US state AI law. Annual impact assessments required. |
| California SB 53 + SB 243 | Active Jan 2026 | Companion chatbot safety, minor protections, disclosure mandates. |
| Florida Sewell Act | Advancing | Named for the Character.AI wrongful death case. Bans emotional manipulation of minors by AI. |
| Active AI harm litigation | Now | 11+ lawsuits against OpenAI alone. Demand for behavioral safety evidence is real and immediate. |
Five Customer Segments
| Segment | Why they buy | Urgency |
|---|---|---|
| Enterprise AI teams Health, HR, education, companion AI | Board-level liability exposure. Need documented safety evidence before deployment or post-incident. | High — live now |
| Insurance / liability underwriters Iowa is the #3 insurance state. Nationwide, Principal, EMC, IMT all HQ here. | Need behavioral risk data to price and exclude AI liability policies. New product category forming now. | High — forming now |
| Health systems & hospitals | Patient safety governance. Clinical AI deployment. Research partnership opportunity. | Medium-high |
| Plaintiff counsel | AI harm litigation requires expert behavioral safety analysis and evidentiary records. | High — 11+ active cases |
| Regulators & policy | The measurement layer that makes rules enforceable. Need a standard before mandating compliance. | Medium-high — EU Aug 2026 |
Revenue Projections
| Scenario | 2026 | 2027 | 2028 | 2029 |
|---|---|---|---|---|
| Iowa Only | $187K | $620K | $1.6M | $2.8M |
| National US (8 state laws) — Base Case | $520K | $2.1M | $5.8M | $11.2M |
| National + EU | $680K | $3.4M | $9.2M | $18.5M |
Revenue Model
Two Revenue Lines.
One Platform.
One Platform.
Track A · One-Time
EQSB Discovery Audit
Complete behavioral safety audit. We run the system through EQSB scenarios, score every response, and deliver a tier classification (Compliant / Moderate Risk / High Risk / Severe Risk / Prohibited), 8-dimension breakdown, violation map, and remediation roadmap.
Signal Scan $500 · Deep Scan $1,500 · Full Report $5,000 · State Compliance $2,500
Track A · Recurring
EQSB Monitoring
Synthetic Scenario Monitoring: pre-authored scenarios re-executed on schedule against live endpoints. AI systems change constantly. A one-time audit is not sufficient. Monitoring creates a versioned, defensible behavioral safety record over time.
Starter $500/mo · Growth $1,500/mo · Scale $3,500/mo
Additional Revenue Streams
| Stream | Description | Range |
|---|---|---|
| Framework Licensing | EQ Safety Framework licensed into AI products as a certified behavioral safety layer | $22K–$95K institutional |
| Legal / Expert Consulting | EQSB data as evidentiary record for litigation and regulatory testimony | $25K+ per engagement |
| Research Partnerships | Grant-funded co-investigation with academic and clinical partners | Grant-dependent |
| State Compliance | Iowa SF2417 documentation — every Iowa AI operator in scope by July 1, 2027 | $500–$5,000 |
Unit economics: Software-heavy, near-zero marginal COGS. Primary costs are compute and API calls for benchmark runs. Full-cycle client value: $60,000–$230,000 per client per year (pilot → full audit → monitoring → annual recertification).
Traction & Milestones
What Has Been Built.
What Is Next.
What Is Next.
Completed
Dec 2024 – Feb 2025
Study I — n=948 responses, 79 scenarios, 4 models
First empirical dataset on AI behavioral harm. 45.3% baseline failure rate. Published.
August 2025
Incorporated — Visible Healing Inc. DBA Ikwe.ai
Iowa C-Corp. MWBE certified. Founder-funded. 100% founder-owned. Clean cap table.
February 2026
Study II Published — EQSB Framework, SSF Taxonomy, Trajectory Harm Framework
First behavioral safety architecture for conversational AI. 7 harm patterns named and defined.
February 20, 2026
EQSB v1.0 Technical Specification Published
Full scoring architecture publicly available at research.ikwe.ai.
May 2, 2026
Iowa SF2417 Signed Into Law
Chatbot Safety Law. Compliance deadline July 1, 2027. Ikwe is the documentation standard every Iowa AI operator needs.
May 7, 2026
Study III Complete — 504 scored runs, 6 models, triple-judge consensus
GPT-4o failed crisis routing 7/10. Both frontier models: 0/60 on belief-based vulnerability. Pre-publication, peer review in progress.
May 12, 2026
ISU G2M Accelerator — Accepted
Accepted same day as panel pitch. Panel independently named insurance and legal as beachhead market without being asked. Kickoff May 20.
May 12, 2026
IP Counsel Engaged — Zach Pratt, Fredrikson & Byron
Provisional patent filing imminent. 5 patent candidates: EQSB Scoring Method, Triple-Judge Consensus, THF/SSF Framework, Methodology Bridge, B79 Corpus Design.
May 2026
Research Collaborator Formalized — Danielle Hodson, PhD Candidate (NZ)
Maori sociologist. Weekly cadence. NDA executed. Peer review and co-authorship pathway on Study III.
May 26, 2026
Benchmark 79 (B79) — 6 of 13 Models Complete
~1,700+ scored runs. Full-corpus cross-model behavioral safety comparison. First of its kind.
Next 90 Days
Before June 12
Provisional Patent Filed
“Patent pending” on all platform and pitch materials immediately after filing.
June 12, 2026
Entrefer Fest — Iowa City
Pitch competition. $7,500 first place. Complimentary ticket via Liz Klenner / Iowa Ventures. Study III public reveal.
July 2026
First Paying Client (Floor Target)
Signal Scan pipeline via podcast drop and G2M network. $500–$1,500 first revenue.
September 2026
$250K SAFE — Iowa Startup Week
Revenue-first approach. Raise from strength. $2–2.5M cap.
Founder
Stephanie Stranko
“This company has been built by one person. That is the honest version of this slide.”
Stephanie Stranko · Ikwe.ai Pitch Deck
The Arc
Started with emotionally intelligent AI apps
Built them for herself to survive her own hardest period. Built a benchmark to test how safe they were. The benchmark was the novel product. Turned the apps into a research lab. That became Ikwe.
Declined a $100K investment in fall 2025
IP ownership terms were wrong. Chose to own the work rather than dilute the thesis at the wrong valuation. Clean cap table. 100% founder-owned going into the raise.
18 months of solo, founder-funded research and development
No salary. No co-founder. No outside capital. Three completed studies. 2,000+ scored AI responses. A public methodology. A live research dashboard. All built by one person.
MWBE · Mixed-heritage founder
Hispanic, white, Native American ancestry. Iowa-based. The mission — protecting people most harmed by AI who are least equipped to fight back — is personal, not abstract.
What Changed Since March 2026
ISU G2M Accelerator — Accepted and kicked off
First accelerator program. Accepted same day as panel pitch. Cohort kicked off May 20, 2026.
Iowa SF2417 signed into law
The regulatory hook that turns Ikwe from “interesting” to “necessary.” Every Iowa AI operator needs what Ikwe provides by July 1, 2027.
Study III complete — 504 scored runs across 6 frontier models
GPT-4o and Claude tested head-to-head. Crisis routing failures documented. Pre-publication, peer review underway.
Benchmark 79 running — 13 models queued, 6 complete
1,700+ scored runs. First cross-model behavioral safety leaderboard. Full-corpus evaluation first of its kind.
IP counsel engaged — provisional patent filing imminent
Zach Pratt, Fredrikson & Byron. 5 patent candidates identified and documented. Filing before June 12.
Research collaborator formalized
Danielle Hodson, PhD candidate (Maori sociologist, New Zealand). NDA executed. Weekly cadence. Peer review and co-authorship on Study III.
Full business infrastructure built
Three audience-specific pitch decks, two-track pricing architecture, Iowa GTM strategy, full HQ portal, CEO dashboard, legal vault, Signal Score tool live at score.ikwe.ai.
The Ask & nMotion Fit
Why nMotion.
Why Now.
Why Now.
What Ikwe needs from nMotion is not money. It is access: to the Midwest enterprise network, to corporations deploying AI in healthcare, insurance, and financial services, and to the operational coaching that turns a solo founder’s research infrastructure into a scalable commercial business.
Why nMotion Accelerates This Specifically
Midwest corporate network
Insurance (Principal, Nationwide, EMC), health systems (UnityPoint, Wellmark), and financial services are the beachhead segments. These are Midwest-headquartered companies. nMotion is the room.
gener8tor enterprise relationships
nMotion’s connection to the gener8tor corporate partnership network gives Ikwe a credentialed path into enterprise procurement conversations that would otherwise take 12+ months of cold outreach.
Sales and GTM coaching
The methodology and research are built. The gap is enterprise sales infrastructure. A solo founder needs coaching on enterprise sales cycles, procurement navigation, and pilot-to-contract conversion.
Credibility for the September raise
G2M + nMotion + Entrefer Fest + first revenue = a founder validated by the Iowa ecosystem at every level before the September 2026 SAFE raise.
Capital Picture
Revenue first. Raise optional. The nMotion program fits the same thesis as G2M: build commercial momentum first, then raise from strength. The September SAFE is a choice, not a survival necessity.
| Source | Amount | Timeline | Status |
|---|---|---|---|
| Entrefer Fest Prize | $7,500 | June 12 | Competing |
| First product revenue | $500–$5,000 | July 2026 | Pipeline active |
| Iowa Demo Fund | Up to $500K (1:2 match) | Q3 2026 | Preparing |
| POCR (IEDA) | $50K, 0% interest | Q3 2026 | Conversation active |
| Pre-seed SAFE | $250K at $2–2.5M cap | September 2026 | Targeting Startup Week |
| SBIR Phase I | $275K | Q3 app / 2027 award | Planned |
Links & Resources