ProfytAI
Regulatory Evidence

The Examiner Is Asking.
Do You Have the Evidence?

A compliance answer is only worth what you can prove under examination.

The Real Question

Can a General-Purpose AI Tool Produce Regulatory Evidence?

Banks ask whether general-purpose AI tools can replace a purpose-built compliance evidence system. This is not a question of capability. It is a question of architecture.

The AnswerNo, It Cannot
The Examiner’s Demand

“Show Me Your Record for This Obligation.”

BSP.MORB.2023.Sec921.d.p22.OBL1 · Face-to-Face Contact · BSP MORB Part IX

General-Purpose AI

It Improvises an Answer

Input · A Compliance Question

01

Searches for Fragments

Finds pieces that seem related, not the full obligation.

02

Generates a New Answer

A different answer every time you ask.

03

Nothing Is Stored

The session ends and it is gone.

04

Cannot Be Traced

No source text, no audit trail.

No Artifact

Session Ends

Dead End

Nothing stored or versioned.

No ReplayNo Trail
ProfytAI Evidence System

It Maintains the Register

Input · The Full Regulation

01

Retrieves the Stored Obligation

Every obligation preserved word-for-word.

02

Returns Verbatim Output

The same obligation ID returns the same text.

03

Stored and Versioned

Persistent, reproducible, retrievable.

04

Fully Traceable

Source text, mapped controls, evidence.

Audit-Ready

Artifact Stored

BSP.MORB.2023.Sec921.d.p22.OBL1 · v1 · p. 22

Versioned, reproducible record.

VerbatimControlsEvidence

These Are Not Two Versions of the Same Tool.

General-Purpose AI is a search and summarisation assistant. ProfytAI is a compliance evidence system. They solve fundamentally different problems.

The Examination Test

Three Questions Every Regulator Asks. General AI Fails All Three.

Standard examination questions. Any compliance tool that cannot answer them is a liability, not an asset.

Examiner Question

General-Purpose AI

Chat Assistants

ProfytAI

Evidence System

Q01

“Show me the exact regulatory text your control satisfies.”

Every examiner asks this first. The answer must be verbatim. Paraphrasing is not acceptable.

Cannot Comply

Paraphrases source text. The regulator's exact words are not preserved.

Verbatim Source Text

Stored word-for-word at ingestion. Retrievable on demand.

Q02

“Reproduce this compliance output exactly as it was generated.”

Reproducibility is a core examination requirement. If the output changes on re-run, it is not evidence.

Cannot Comply

Probabilistic by design. Every run produces a different output.

Reproducible on Record

Once generated, the output is pinned as a versioned record and returns identically on every retrieval.

Q03

“Provide the audit record and evidence chain for this finding.”

Without a stored record, there is no audit trail. Without an audit trail, there is no compliance.

Cannot Comply

Session-based. When the session ends, nothing is stored. No record exists.

Versioned Artifact

Every output is stored, versioned, and retrievable. Always.

0 / 3

General AI · Failed

3 / 3

ProfytAI · Passed

Verdict of the Test

General AI was designed to answer questions. ProfytAI was designed to produce evidence that withstands examination. These are not the same system.

Why General AI Fails Under Examination

One Examination. Three Architectural Failures.

These are properties of how language models work. No model, no prompt, and no configuration removes them.

Exhibit
Fig. 03a
SAME QUERYEVERY TIMERUN 1are expected to maintainRUN 2should implementRUN 3must establish≠≠
Outputs Diverged · No Canonical Record

Failure 01

Non-Reproducibility

Same query. Different output. Every time.

Under Examination

The examiner asks you to reproduce your evidence.

What Happens

The output will differ. A record that cannot be reproduced on request is not a record.

Fixable by Configuring the AI Differently? No.

Language models sample from probability distributions. Even at temperature zero, outputs vary across sessions and model versions. There is no mechanism that guarantees identical output.

Exhibit
Fig. 03b
SOURCE · MAS TRM § 7.7.4 · P.26 · VERBATIMMODEL OUTPUT · PARAPHRASEAI SYNTHESIS“The FI should configure system events or alertsto provide an early indication of issues…”“Financial institutions are expected to configureappropriate monitoring mechanisms to ensureearly detection of system issues.”
Obligation Reworded · Legal Precision Lost

Failure 02

Paraphrasing

The regulator's words, rephrased by a model.

Under Examination

The examiner asks for the exact words of the obligation your control satisfies.

What Happens

The AI output does not contain them. A paraphrase of a statute is not that statute.

Fixable by Configuring the AI Differently? No.

Language models generate text. Generation means producing new words. The original regulatory text is lost in the process, even with retrieval-augmented generation.

Exhibit
Fig. 03c
MODEL OUTPUT · INFERRED CITATION“Refer to MAS TRM, Part IV (data management section)for the applicable requirements.”CITATIONGENERATEDPREDICTIONNOTHINGTO PRODUCECITATIONRETRIEVEDRETRIEVALSOURCE TEXTSTORED RECORD
Evidence Chain · Broken at Retrieval

Failure 03

Inferred Citation

A reference with nowhere to go.

Under Examination

You cite the obligation. The examiner asks for the source text your system retrieved.

What Happens

You cannot produce it. The citation was generated, not retrieved.

Fixable by Configuring the AI Differently? No.

The AI infers what the citation should be based on patterns in its training data. It does not retrieve a verified source record. The reference is a prediction, not a retrieval.

Not a Model Problem

A better model will not fix this. These failures come from how every language model works. Evidence needs a different architecture.

The ProfytAI Standard

Examination-Grade Compliance. Built to Be Defended.

PAGE 26VERBATIM · LOCKED TO PAGE“The FI should configure systemevents or alerts to provide anearly indication of issues…”MAS TRM · § 7.7.4 · PAGE 26
CAP 01

Verbatim

Source Text Preserved

Every obligation stored exactly as written in the regulation. No paraphrasing. No summarization. The regulator's exact words, retrievable on demand.

STORED RECORD · V1MAS.TRM.2021.Sec7.7.4.p26.OBL1RETRIEVAL 01SAME RECORDRETRIEVAL 02SAME RECORDRETRIEVAL 03SAME RECORD
CAP 02

Reproducible

Same Output Every Time

Any compliance artifact can be reproduced identically at any point in time. The output is the same output that was produced at ingestion.

CITATION CHAIN · MAS TRM § 7.7.4OBLIGATIONSTOREDCLAUSESOURCECONTROLMAPPEDEVIDENCEATTACHEDEVERY LINK VERIFIED AND RETRIEVABLE
CAP 03

Traceable

Every Citation Verified

Every obligation links directly to its source clause. Every control maps to the obligation it satisfies. Every evidence item is attached and retrievable.

The Four Guarantees

Guarantee 01

Every obligation is traceable to its exact source clause

Guarantee 02

Every output is reproducible on demand

Guarantee 03

Every artifact is stored, versioned, and retrievable

Guarantee 04

Every citation is independently verifiable

These properties apply to content and workflows delivered through the ProfytAI platform for covered regulatory material. They do not replace your policies or sign-offs.

A language model produces answers. An evidence system produces proof. The difference is not quality. It is architecture.

What ProfytAI Is

ProfytAI is not a compliance assistant. It is a compliance evidence system, the layer between your institution and your regulator that makes examination a process of retrieval, not reconstruction.

How the Engine Uses AI

AI Proposes. The Source Decides.

Our compiler uses AI to propose each record. Verification against the source text decides what is kept, and the Integrity Ledger stores it.

Can a Better Prompt or a Newer Model Fix This?

No. Language models sample from probability distributions, so output can vary across sessions and model versions, even at temperature zero. Generation produces new words, so the regulator's text can be lost even with retrieval-augmented generation. A generated citation is a prediction of where the rule should be, not a retrieval of a stored record. Changing models does not change these properties.

PROPOSALACCEPTEDAI PROPOSESPROPOSED RECORDVERIFICATIONAGAINST THE SOURCE TEXTINTEGRITY LEDGERSTOREDSTOREDSTOREDREJECTEDNOT KEPT

See the Evidence
for Yourself.

We will walk through a live compliance artifact. Obligation ID, verbatim source text, evidence chain, and audit record.