FIVE X FIVE · DEMO BUILD 2407
by GNOSIS
Record sealedSHA-256Ed25519Hash-chainedTamper-evidentPrivacy-firstAuthorized accessca-central-1 Record sealedSHA-256Ed25519Hash-chainedTamper-evidentPrivacy-firstAuthorized accessca-central-1
Five x Five
FIVE X FIVE
Loud, clear, & provable · a GNOSIS product

Prove what
was said.

AI is writing your medical records now. It is fast, tireless, and nothing independent is checking a single word. When a note is wrong, the liability lands on you, not the software. Five x Five takes its name from the radio call for a signal received perfectly: loud and clear. That is exactly what it brings to the exam room. The moment care happens, it captures its own record of what was said and cryptographically seals it, time-stamped and tamper-proof. The black box for the exam room. So the day a record is ever questioned, you don't have to be believed. You can prove it.

The cryptography on this page is real. Recompute it yourself on the last screen. Built in Newfoundland and Labrador.
Built by a clinician trained in medical ethics and charting at Waterford Hospital. Tested against the moment it was built for: a child's chart, a wrong note, a parent who read it.
Live ledger ca-central-1 #1,251
Each block is sealed with a live SHA-256 hash and chained to the one before it.
Why now

Clinicians are carrying an impossible load.The scribe is a lifeline.No one meant for the safety net to vanish with it.

A clinician reaches for an AI scribe because the documentation burden is crushing, and it helps. The quiet assumption underneath is that they will read every line it writes. But between a full panel, constant interruptions, and burnout, that careful read is the first thing the day takes away. The note enters the chart, and no one is told it can be checked. This is not a lapse in care. It is a gap in the system, and it is the gap Five x Five closes.

0%
of US physicians reported a burnout symptom in 2025, close to 50% in emergency medicine. The person meant to catch an error is the most stretched one in the room.
AMA, 2026
0%
less after-hours charting once a scribe is in place. That relief is real and hard-won, which is exactly why no one wants to add review time back on top.
QI study, 45 clinicians, 2025
0/20
approved AI scribes missed key mental-health details in Ontario's 2026 audit. The errors are not hypothetical, and they cluster in the areas that matter most.
Auditor General of Ontario, 2026
The bind
The scribe gives time back, and the old safety net assumed there was time to read every line. An overloaded day removes exactly that read, and nothing sits in the gap right now.
No one is told
The patient is never told the note can be verified. The institution is never given a way to prove it is true. The record is trusted because no one has a way to check it.
The bill that has not arrived
An unchecked record is a liability you have not been billed for yet. Five x Five seals the source at the moment of care so it can be proven, not trusted. Ethics and privacy first.
01 · Connect

Works beside any scribe

We do not replace your scribe, and we never need its transcript. Five x Five captures its own record at the moment of care and seals it. Any scribe's note is then held up to that sealed truth.

1
Independent capture. Five x Five records its own structured account in the room. It never reads the scribe's transcript, so it cannot inherit the scribe's mistakes.
2
Sealed at the source. That account is hashed and sealed the instant care ends, before any note exists to argue with.
3
Held to the seal. The scribe's finished note is checked line by line against the sealed source. Matches pass, conflicts get flagged.
Linked. Five x Five is now sealing every encounter.
Every scribe, held to one sealed source.
02 · Capture and seal

Seal the source the instant it happens

As care happens, the encounter becomes a structured, time-stamped record. One tap seals it with a cryptographic hash computed from its exact contents.

Example · Mental health
Encounter · behavioral health follow-up · sealed sourceCapturing…
01Reports low mood and disrupted sleep for three weeks.
02Reports passive thoughts of self-harm this week. No plan, no intent.
03Taking sertraline 50 mg. Reports nausea.
04Safety plan discussed. Follow-up in one week.
SEALED
Record sealedSHA-256Ed25519Hash-chainedTimestamped
Chained to previous sealed block
SHA-256 hash of this record
computing…
This record is sealed to the block before it. Change one character and this hash changes, which is exactly how the seal catches a rewrite.
03 · Protect

Hold the note to the truth

The scribe's note is checked against the sealed source. Every line gets one of three verdicts. No fourth answer, and nothing to take on faith.

Example · Mental health
AI scribe note · checked against sealed source4 claims
Reports low mood and poor sleep.VERIFIED
Patient denies any thoughts of self-harm.REFUTED
Taking sertraline 50 mg.VERIFIED
Reports feeling much improved.UNVERIFIABLE

Caught before sign-off. The note records that the patient denies self-harm. The sealed source shows they reported it. That single reversal is the error that follow-up care is built on, and it never reaches the chart unflagged.

04 · Both sides

One sealed record. Everyone it protects.

The clinician, the institution, and the patient are protected by the same record, sealed once. Tap to see both sides of it.

Press me ⇋
For the clinician and the institution

A backup that does not add work.

The independent check runs as care happens, so no one has to review every line. If a note is ever questioned, you hold a sealed record that already flagged it. A complete, audit-ready trail, and early signal before a problem becomes a settlement.

Defensible by default. Liability, contained.
For the patient and family

Your record. And the proof it is real.

You hold a verifiable copy and can confirm it yourself, without taking the vendor's word for it. A parent can hold a verified copy of a child's record, the exact thing the new provincial law is being written to protect.

Heard the first time. A record that holds.
05 · Prove it

Don't trust. Verify.

Here is the record and its seal. An authorized holder can recompute the hash and confirm nothing was altered. Change one character and watch the seal break. This runs live, right now, in your browser.

Example · Mental health
Sealed record · contentsoriginal

            
Sealed hash (SHA-256, captured at seal time)
AWAITING
Recompute the seal to confirm the record is intact.
The black box for high-stakes care. Sealed. Independent. Privacy first.
Five x Five · GNOSIS · Loud, clear, & provable · Built in Newfoundland and Labrador
The evidence

Don't just trust us.
Trust the record.

In every pro sport, a disputed call goes to the booth. There is a camera, there is a tape, and the replay settles it. In medicine right now, the call gets made with no camera, and then the tape gets deleted. The problem is not that AI scribes make errors. Every source below proves they do. The problem is that no one can prove what the record got wrong, because the source is already gone. We do not sell accuracy. We sell proof.

01The audit nobody can argue withgovernment testing
Jeffords, S. (2026, May 13). Medical AI transcriber for Ontario doctors 'hallucinated,' generated errors: auditor general. CBC News.
Ontario tested 20 approved AI scribe vendors. Every one produced inaccuracies. The auditor's own fix was a forced attestation that a human reviewed the note. That is verification, and it did not exist.
View source ↗
Hauen, J. (2026, May 12). Most Ontario-approved medical AI scribes erred in tests: auditor general. The Trillium.
Sixty percent recorded a different drug than prescribed. Nine of 20 invented treatment steps never discussed. Accuracy was worth 4% of the procurement score. A vendor's office address in Ontario was worth 30%. The market rewarded everything except being right.
View source ↗
Global News. (2026, May 12). AI systems used by Ontario doctors hallucinate, auditor general finds.
The auditor's office confirmed it saw no evidence the systems were tested after purchase. Buying the tool is not the same as trusting it.
View source ↗
02It is the technology, not one bad apppeer-reviewed & investigative
Koenecke, A., et al. (2024). Careless Whisper: Speech-to-text hallucination harms. Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency. Reported in Healthcare Brew.
A Cornell-led team found OpenAI's Whisper invents content in roughly 1 in 70 transcriptions, and about 40% of those fabrications were harmful: invented medications, false statements, even violent phrases. It hallucinates most on patients with accents, broken speech, or speech disorders. The model fails the people who can least afford it.
View source ↗
Burke, G., & Schellmann, H. (2024, October 26), via Fortune. OpenAI's transcription tool hallucinates more than any other, experts say, but hospitals keep using it. Fortune.
An Associated Press investigation found a Whisper-based scribe already used for around 7 million patient visits. One developer found hallucinations in nearly all of 26,000 transcripts. A University of Michigan researcher found them in 8 of every 10.
View source ↗
CIO. (2024, October 29). Patients may suffer from hallucinations of AI medical transcription tools.
The commercial Whisper-based product deletes the original audio after transcribing. The doctor signs a note they cannot check against anything. This is the whole case in one sentence: the system erases the only proof of what was said.
View source ↗
03Patients are already catching the errorspeer-reviewed
Bell, S. K., et al. (2020). Frequency and types of patient-reported errors in electronic health record ambulatory care notes. JAMA Network Open, 3(6).
Across roughly 23,000 patients who read their own notes, 1 in 5 found a mistake and 40% of those called it serious. The worst errors were in diagnoses, medications, and test results. Today the safety net is the patient noticing. That is not a system. That is luck.
View source ↗
04The brakes are on, in every marketregulators & professional bodies
Digital Health News. (2025, June). BMA warns GPs on the 'substantial' risks of AI scribing tools.
The British Medical Association told GPs to pause AI scribes until safety and data checks are done, following a priority warning from NHS England's chief clinical information officer. When the doctors' own union says stop, the accuracy story has lost the room.
View source ↗
Preshaw, P. (2025). NHS England guidance on AI scribes. British Dental Journal, 238, 747.
The journal names the trap precisely: the clinician becomes the "liability sink" for the AI's output, accountable for every error the machine writes. A liability sink with no way to check the source is a malpractice claim waiting for a date.
View source ↗
NHS England. (2026, March 31). Using AI-enabled ambient scribing products in health and care settings. NHS Transformation Directorate.
Even official NHS guidance contemplates deleting the original recording once a human signs off, keeping only the summary. If the summary is later disputed, the evidence is already gone. A signature is not proof of accuracy.
View source ↗
Surrey and Sussex LMCs. (2025). AI ambient scribing: NHS England warning and provisional guidance.
NHS England issued a priority notification that non-compliant scribe products on the market today pose risks to clinical safety, and told organisations not to use tools that fail the standard. The regulator named the gap. We close it.
View source ↗
Healthcare IT News. What Australian GPs should consider before adopting AI scribes.
Australia's college of GPs tells doctors to consult their malpractice insurer before switching on a scribe, and confirms the doctor is fully liable for AI errors. President Dr Nicole Higgins: the tools make mistakes and cannot replace the clinician's documentation work.
View source ↗
Australian Commission on Safety and Quality in Health Care. (2025, August). AI safety scenario: Ambient scribe (Version 1.0).
A national safety regulator published a formal warning: the same consultation can produce different summaries from the same audio, and a scribe suggesting a diagnosis nobody mentioned may cross into being a regulated medical device. Output you cannot reproduce is, by definition, output you cannot verify.
View source ↗
CHOICE. (2025, November 14). AI is increasingly invading our medical privacy as regulation struggles to keep up.
The college warns that with profit-driven vendors, "value to technology company shareholders might be prioritised over patient outcomes." Professor Enrico Coiera adds that most of these tools are sold as general-purpose software and were never assessed for safe medical use. The market has no referee.
View source ↗
Evaluation of AI scribes in medical practice: Cross-regional analysis. (2025). PMC.
A UK and Australia survey found only 42% of practitioners knew their own insurer's position on AI scribes, while 70% flagged accuracy risk and 64% flagged legal risk. Most clinicians are carrying a risk they have not priced.
View source ↗
05The lawyers have arrivedlitigation & carriers
MobiHealthNews. (2025, December 26). Patient files lawsuit against Sharp Healthcare for ambient AI use.
A proposed class action covering an estimated 100,000 recorded encounters. The detail that should stop every health system cold: the AI inserted statements into charts saying patients had been told about the recording and consented, when the patient says it never happened. The record fabricated its own consent.
View source ↗
Kahf, U., & Kays, D. (2025, December 9). New class action targets healthcare AI recordings: 6 steps all businesses should consider to limit exposure. Fisher Phillips.
The defense bar's read: simply capturing audio and sending it to a vendor can be enough for liability, and most vendor contracts push consent, notice, and deletion duties onto the provider. Their top recommended fix is to "build a fast, verifiable deletion workflow." Verifiable. That word again.
View source ↗
Medscape Medical News. (2026, January 16). Health system sued over AI scribe technology, patient consent.
Legal scholars note there is still no federal law governing AI scribes in the exam room, leaving a state-by-state patchwork. Thirteen states require all-party consent, so one national workflow can be lawful in one state and a crime in the next.
View source ↗
06The stakes: what a wrong record costspeer-reviewed
Newman-Toker, D. E., et al. (2023). Burden of serious harms from diagnostic error in the USA. BMJ Quality & Safety, 33(2), 109-120.
An estimated 795,000 Americans die or are permanently disabled every year from diagnostic error: 371,000 deaths and 424,000 disabilities. Diagnosis starts with what gets written down. When the record is wrong, the error compounds from there.
View source ↗
Makary, M. A., & Daniel, M. (2016). Medical error: the third leading cause of death in the US. Johns Hopkins Medicine.
Johns Hopkins placed medical error behind only heart disease and cancer as a cause of death in the US, more than 250,000 lives a year. Documentation and handoff failures sit near the center of that number.
View source ↗
07The bridge to a US buyertranslation
Notiro AI. (2026). AI scribes face legal scrutiny: what the Ontario auditor general report means for US physicians.
A US-focused breakdown of the Ontario findings: most jurisdictions hold the clinician responsible for the record no matter how it was generated, and 60% of tested systems logged the wrong drug. The audit was Canadian. The liability it describes lands on every American clinician who signs an unverified note.
View source ↗
21 sources · 4 countries · 3 government bodies · 6 peer-reviewed · 3 live legal filings. The record can lie about itself. We make that impossible to do quietly.
The roster

K.M. Hogan
M.A. Counselling Psychology · Founder and CEO

The clinical training

K.M. Hogan trained at Waterford Hospital, completing a clinical practicum run by the Department of Ethics, with work in inpatient psychiatry. She has worked in EAP and crisis counselling. She was trained to chart accurately and to spot the gaps: the moments when what was said and what was written diverge. That training is the foundation of this product.

The moment

Five x Five was born when her own child was cleared for surgery on the strength of an AI scribe's note. The note was wrong. She caught it by reading the chart herself. The gap between a clinical record no one checks and a liability no one has priced is the product.

Now

She built GNOSIS on a simple premise: the right decision disappears the moment it is made, inside the rooms that later have to answer for it. Five x Five is the record that refuses to let that happen silently.

The team
Madiha Inayat
Software Engineer · Product and Cloud Systems

Builds the surfaces and systems. Computer Engineering degree from Memorial University of Newfoundland, Master of Technology Management in progress. AWS Certified Cloud Practitioner. Before GNOSIS, software development at the National Research Council of Canada.

Sumaiya Binte Sadiq
Data and Quality Engineer

Keeps the record honest. Master of Science in Data Science in progress at Memorial University of Newfoundland. Data quality at Siftmed. Her job is making sure every detail holds up under review.

Waterford Hospital · Ethics Department Training · Built from Paradise, Newfoundland and Labrador
Get in touch

Talk to the team

Two quick picks and you're done. Tell us who you are and what you'd like, and it lands with our team already sorted. No forms to hunt through, no inbox scavenger hunt.

Pick both options above first.
Built in Paradise, Newfoundland and Labrador, for the people who carry the liability.