We catch AI hallucinations

before they happen — and prove it.

The architecture-portable internal-state control plane for AI reliability.

Seed Financing · $6M SAFE · $30M Post-Money Cap · 20% Discount

EXECUTIVE SUMMARY

Today's AI Is Flying Blind

Enterprises are spending $407B on AI in 2026 alone1, 2 — and accelerating

AI errors and hallucinations are already producing real, escalating financial and legal exposure — court sanctions in the hundreds of thousands of dollars, McKinsey citing security and risk as the #1 barrier to further deployment3–10

The Current Approach Falls Short: Most organizations still catch AI errors only after output — checking, correcting, and redoing work after the fact — which is slow, costly, and too late for high-stakes use cases

Solution: Proprioceptive AI V1 reads a model's internal activity in real time, flags likely errors before they reach the output, applies a correction and verifies the result — reporting an 85.8% reduction in confident-wrong output in internal testing. V2 expands this architecture across additional models, workflows and enterprise deployments11

The Ask: $6M to scale V1, accelerate V2, initial enterprise pilots, early commercialization and focused patent prosecution

2

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

Proprioceptive AI will fund up to $50K toward independent technical due diligence — investor's reviewer, investor's terms.

THE MARKET

Billions in AI Spend. A Real, Costly Error Problem.

Proprioceptive AI is focused on enterprise and government deployments — finance, healthcare, legal, and regulated SaaS — where an AI error carries financial, legal, or safety consequences.

$407B1,2

Global enterprise AI spending in 2026, up 34.8% YoY

4–6%8

Baseline hallucination rate, general-purpose frontier models — rising to 15–40% in open-ended and multi-step agent tasks

2 of 3

organizations cite security/risk — not regulation — as the #1 barrier to scaling AI (McKinsey)9

Financial Services

$68B

spend, 79% adoption2

AI trading, fraud, and advisory tools carry real regulatory exposure when they err.

Healthcare

1.47%

clinical hallucination rate, peer-reviewed7

Even at low rates, a fabricated clinical detail can propagate into patient harm.

Legal & Government

1,300+

court cases; sanctions to $109K+5

Even paid, specialized legal-AI tools still hallucinate on real filings.

Enterprise SaaS

Liable

Air Canada precedent6

Courts have ruled: companies own what their AI says — “separate entity” defenses fail.

Enterprises are spending hundreds of billions on AI while McKinsey finds most are not yet governance-ready for the errors that follow.

Reducing that error rate is the return this category is built to capture.

3

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

WHAT WE DO

Find. Fix. Prove.

Many evaluations inspect only what a model says. Proprioceptive AI V1 also measures internal-state patterns linked to behavior — closer to an MRI for an AI system than a surface-level filter — then corrects and verifies the result.

01

Find

Trained behavioral probes read selected hidden states and flag patterns associated with hallucination, deception, manipulation and sycophancy before output.11

AUC 0.96–0.999

under defined test conditions (= near-perfect on 0.5-1.0 scale)

02

Fix

Trained correctors reduce targeted failure modes, while a dedicated capability probe checks for regressions. No full retraining from scratch.11

85.8%

reduction in confident-wrong output

03

Prove

You receive a cryptographically signed, capability-checked certificate documenting the measured outcome — evidence suitable for technical review, customer diligence or governance workflows.

Dual-Ed25519

signed certificate

4

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

On-prem deployment · One framework across architectures · Zero-shot transfer in validated settings ·

Runtime gate for policy-disallowed actions · Honest by design — even nulls are signed

HOW IT WORKS

From Hidden States to a Signed Verdict

Our V1 process, step by step. Deploy on your hardware with results returned as cryptographically signed evidence. V2 expands portability, automation and enterprise deployment scale.

01

Read the Model

Capture internal activity from your AI model — architectures, frozen weights

→

02

Standardize the Signal

Convert into a common, comparable format across model types

→

03

Compress It

Reduce to a manageable size without losing the relevant signal (JL-512)

→

04

Detect Patterns

Trained detectors flag hallucination- and deception-linked patterns

→

05

Issue a Verdict

Return a cryptographically signed certificate of the result

Then control it — in real time

Steer

Nudge activations toward the behavior you want, mid-generation.

Edit

Apply bounded suppression or amplification without full retraining.

Fine-tune

Bake corrections back into the weights when ready.

Gate

Hold policy-disallowed actions until internal evidence checks pass.

5

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

BENCHMARK RECORD11

Headline Results

44

Working Patent Families — provisional filings; patent counsel confirming legal family structure and prosecution priority12

40

Model Types Tested

85.8%

Reported Hallucination Reduction

1,376×

Signal Strength vs. Random Chance

6

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

IP ARCHITECTURE – 44 WORKING INVENTION FAMILIES12

Three Coordinated Crown-Jewel Families

A parent machine architecture, a portable interface across heterogeneous models, and a predictive evidence-governed improvement loop — separated to preserve distinct prosecution and continuation paths.

Family I · Parent Architecture

Dual-Path Neural Computation with a Proprioceptive Operational-State Pathway

A primary pathway produces text, predictions, plans or actions. A functionally distinguishable proprioceptive pathway represents the system's own operational condition while computation is ongoing.

Family II · Interoperability Layer

Architecture-Universal Operational-State Interface

Target-relative coordinate construction, ambiguity stabilization, portable read/write operators, target-native lifting and heterogeneous operational-state packets.

Family III · Governed Autonomy

Proprioceptive Predictive Control and Verified Self-Improvement

Prediction before action, immutable counterfactual records, bounded candidate compilation, isolated execution, protected evaluation, promotion, rollback.

7

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

INDEPENDENT MULTI-LAB CONVERGENCE

The Field Is Moving in This Direction.

Proprioceptive AI Started Early and is Assembling It End-to-End.

Different teams, models and methods are independently exploring pieces of this problem — evidence the direction is real, not evidence any one of them has built the integrated system.

26 Feb 2026

Semantic Tube Prediction

Huang · LeCun · Balestriero16

STP treats token sequences as geodesic trajectories, separating geodesic-parallel signal from perpendicular deviation.

Convergence: Supports the view that hidden-state geometry can be functionally meaningful and actionable.

Boundary: A training objective over token trajectories — not a replication of Proprioceptor's channels or certification.

6 May 2026

Manifold Steering

Wurgaft et al. · Goodfire17

Shows interventions following learned activation manifolds generate more natural behavioral trajectories than Euclidean steering.

Convergence: Supports geometry-aware internal control rather than single Euclidean directions.

Boundary: Controlled manifold experiments — not the full read–act–verify control plane.

6 Jul 2026

Anthropic J-space

A Global Workspace in Language Models15

Identifies a silent internal workspace, readable before output, causally editable, used for multi-step reasoning.

Convergence: Supports consequential pre-output internal state that can be monitored and influenced.

Boundary: A Jacobian-lens workspace on verbalizable concepts — not model-independent instrumentation.

25 May 2026

Google ScientistOne

Chain-of-Evidence · CoE Integrity Audit18

Autonomous research with claim-level evidence chains, auditing reproduction and method-code alignment.

Convergence: Supports the governance principle that autonomy must remain bound to reproducible evidence.

Boundary: Validates the evidence-governance layer — not the hidden-state sensing itself.

Other labs validate individual pieces — geometry, internal workspaces and evidence chains. Proprioceptive AI integrates sensing, intervention and verification into one deployable control plane. The convergence validates the direction; the integration is the product.

8

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

MANAGEMENT & RESEARCH

Logan Matthew Napolitano

Founder & Chief Executive Officer

Logan is a self-taught AI researcher who independently studied interpretability and model internals — including learning Russian specifically to access untranslated research relevant to his work — before founding Proprioceptive AI.

He built the Company's initial probe architecture independently, has authored the technical papers below, and leads the Company's patent strategy with outside counsel.

Nicholas D. Goodman, Managing Director and Chief of Operations, leads day-to-day operating execution and external workstreams. Logan is also supported by outside legal counsel, a commercial and fundraising advisor, and additional operating contributors as the Company scales.

Published Research (Zenodo)

Self-Compounding Behavioral Probes for Autonomous Model Improvement and Probe-Guided Compression

82 pp.

Consistency Is All You Need

43 pp.

Unified Behavioral Modulation in LLMs: Cross-Architecture Validation of Geometric Behavioral Subspaces

32 pp.

Controlled Language Models: Inference-Time Control, Tokenization Engineering, and Reversible Optimization

69 pp.

Mathematics Is All You Need

459 pp.

5

Published Papers

44

Working Patent Families12

40

Model Types Tested

9

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

COMPETITIVE LANDSCAPE

Proprioceptive AI Positioning

Position By Method Category — Who Uses What Today

Output filters / LLM judges / RLHF

Fast or broad, but generally act after generation; costly to repeat at scale.

Standard today — labs and enterprises

Activation monitoring & interpretability research

Powerful causal tools, but typically model-specific; sensing, correction and rollback rarely integrated.

Labs: internal research. Enterprises: rare, emerging

Proprioceptive AI

Pre-output sensing, targeted intervention, remeasurement and signed evidence — integrated, architecture-portable.

Emerging category — no company has scaled this widely yet

Position By Company14

Company

Founded

Raised / Stage

Last Valuation

What They Do

Goodfire

2024

$209M (Seed→Series B)

$1.25B (Feb 2026)

Mechanistic interpretability; hallucination/deception detection

LMArena

2023 (as Chatbot Arena)

$250M (“Seed”→Series A)

$1.7B (Jan 2026)

Model evaluation & benchmarking

Virtue AI

2024

$30M (Seed+Series A)

Acquired by Fortinet, Aug 2026

AI red-teaming, guardrails, enterprise security

HiddenLayer

2022

$56M raised

Series A stage

ML model security & threat detection

Proprioceptive AI

2026

$500K to date; raising $6M

This raise

Pre-output internal-state sensing, bounded intervention, verified evidence

Proprioceptive AI's methodology is the newest, least-adopted approach in this table — an emerging category, not a crowded one.

10

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

GO-TO-MARKET & BUSINESS MODEL

How This Becomes Revenue

A land-and-expand enterprise motion: paid diagnostics build evidence and reference accounts, which convert into recurring platform agreements. V1 is the wedge; V2 is the expansion layer.

Target Pricing Ladder

Stage

Offer

Indicative Range

Purpose

Diagnose

Model Diagnostic / Internal State Mapping

$25K – $125K

Fast, bounded paid engagement producing an evidence report

Correct

Correction Validation

$125K – $250K

Test one bounded intervention against a no-action baseline

Govern

Private Deployment Evaluation

$250K – $500K

Customer-controlled evaluation with security acceptance criteria

Expand

Enterprise Platform (annual)

$150K – $500K/yr

Recurring software, governance, evidence and support

Strategic

Strategic Model Program

$750K – $4M+/yr

Multi-model, multi-year development and validation program

Who Buys First, and Why

AI Labs & Model Builders

Need to understand and test behavior inside their own supported models before deployment.

Private AI Operators

Need to evaluate model behavior without exposing sensitive workflows to a third party.

AI Safety & Research Groups

Need transparent, reproducible studies — a natural fit for the Company's published, open methodology.

11

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

MILESTONES & USE OF PROCEEDS

What's Built. What This Raise Funds.

Built to Date

V1 built and operational: configured private AI workspace for pilots11

Core probe architecture validated across 40 model architectures11

5 published technical papers, open for review

44 working patent families; ~17 strongest prosecution candidates12

$500K raised to date from 5 individual investors

Next 90 Days

Independent theoretical review and lead-investor technical diligence offer

Package diagnostic/pilot scope, pricing and statement of work

Seal repeatable V1 install, demo and acceptance test

The Raise

$6M · SAFE · $30M post-money cap · 20% discount

Illustrative Use of Proceeds ($6M)

Product, Validation & Infrastructure

$1.5M

25%

Team & Technical Hiring

$1.5M

25%

Commercialization & Pilots

$2.0M

33%

IP, Legal, Compliance, Capital Raise & Working Capital

$1.0M

17%

Roadmap (12 months)

0–3 mo — Validate: complete review, secure 2–4 paid diagnostics

3–8 mo — Prove: convert first pilot to an annual customer

8–12 mo — Scale: 1–3 annual enterprise customers, expand into V2

12

PROPRIOCEPTIVE AI

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

FAQ

Questions, Answered

What exactly is “proprioceptive AI”?

Proprioception is your body's ability to sense its own position without looking. We built small learned monitors (probes) that read hidden states and estimate behavior-linked patterns before output — an external monitor, not consciousness.

How does this compare to RLHF, and can I use my own models?

RLHF modifies weights and can require substantial retraining. Proprioceptive AI can leave the base model frozen and apply targeted interventions at decode time. It is architecture-portable in validated settings across 40 model types from 410M to 405B parameters, including transformers, Mamba, MoE and RWKV.

Can it prevent bad actions — not just detect them?

In supported deployments, detection runs before output, so a response can be held before display and an agent's tool call can remain pending until internal evidence checks pass. The gate can approve, deny or route the action.

What does “1,376× signal strength” actually mean?

It compares our detector's signal to a random-chance baseline under the documented internal protocol — a strong result in those experiments, not a universal multiplier of model quality, and not yet independently verified.

Do we have priority over where AI architecture is heading?

Yes. Proprioceptive AI has built a broad priority position across the integrated model-internal control stack: pre-output sensing, targeted intervention, post-action verification, architecture portability and cryptographically signed evidence. The portfolio spans 44 working invention families, with approximately 17 leading prosecution targets12.

What's the current stage of the company?

V1 is built and operational. V2 is under active development to expand portability, automation and enterprise deployment scale. The Company is pre-revenue, with internally validated technology, 44 working patent families and cross-architecture results. No customer pilots, LOIs or commercial deployments are signed to date — see Milestones and Risk Factors.

13

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

RISK FACTORS

Specific Risks

For institutional investors only; not for retail investors.

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

Company & Commercial

Pre-revenue; no operating history

No signed customers or LOIs to date

Unit economics, pricing, and sales cycle unproven

Single-founder dependency for core technical knowledge

Limited operating team; key hires still pending

Technical

Performance metrics are internally generated and not yet independently verified

Results may not reproduce across models, tasks, or environments

Independent technical review not yet complete

Some published research is exploratory and outside core validated claims

Intellectual Property

Patents are pending, provisional filings only — none issued

Priority dates, claim scope, and enablement not yet tested

Freedom-to-operate not independently assessed

Prosecution outcomes and timing uncertain

Structural & Financing

Financing structured as a SAFE; converts to equity at a future priced round

Additional capital will be required beyond this raise

Integration depends on continued access to third-party model internals

Regulatory environment (e.g., EU AI Act) is still developing

Private securities are speculative, illiquid, and carry a high degree of risk — including the loss of the entire investment

14

RISK FACTORS

General Risks

Financial models and other supporting information regarding historical data, hypothetical target returns, contextual analysis, and other pertinent matters will be made available to prospective investors upon request. There is no guarantee of success, and there is a potential for loss of your investment.

15

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

Risks related to market conditions and governmental programs

We may be adversely impacted by weakness in the local economies we serve New legislation and/or regulatory laws affecting our operations may affect our performance

Risks related to our business / operations

We may be unable to successfully execute and manage our growth strategy Our success will depend on our ability to hire, train and retain key personnel

Risks related to our industry

The industries in which we operate are highly competitive and many of our competitors have access to greater financial resources, lower funding costs and greater access to liquidity Unfavorable future conditions could adversely impact our business, financial position, results of operations and/or cash flows

Risks related to funding sources and interest rates / debt markets

We are substantially dependent upon our secured and unsecured funding arrangements. If any are terminated, not renewed or otherwise become unavailable to us, we may be unable to find replacement financing on economically viable terms, if at all, which would have a material adverse effect on our business, financial position, results of operations and cash flows Changes in economic and legal conditions could materially and adversely affect our transactions, business, financial position, results of operations or cash flows Our earnings and reputation may be adversely affected if risk is not properly managed

Risks related to proposed financing transaction

There is no assurance that we will be able to obtain additional capital as needed There is no guarantee of success, and there is a potential for loss of your investment There can be no assurance that any financial projections delivered to a prospective investor will accurately reflect the predicted profitability of our business There is currently no market for our securities; and it is unlikely that a market will develop in the future, which would impact your ability to resell our securities in an orderly fashion

SOURCES

Cited References

Numbers in brackets throughout this presentation refer to the sources below.

1

Gartner, "Forecasts Worldwide AI Spending to Grow 47% in 2026," May 19, 2026 — $2.59T forecast for worldwide AI spending.

3

Dahl, M., Magesh, V., Suzgun, M., Ho, D.E., "Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models," Stanford, 2024 (800,000+ verified legal queries).

5

Charlotin, D., "AI Hallucination Cases Database," HEC Paris (ongoing); ComplexDiscovery tracking of per-matter sanctions escalating from ~$5,000 (2023) to $109,700 (early 2026); 1,598 documented cases worldwide as of June 2026.

7

Asgari, E., Montaña-Brown, N., Dubois, M., et al., "A Framework to Assess Clinical Safety and Hallucination Rates of LLMs for Medical Text Summarisation," npj Digital Medicine 8, no. 274 (2025) — peer-reviewed, 12,999 clinician-annotated sentences.

9

McKinsey & Company, "State of AI Trust in 2026: Shifting to the Agentic Era" (2026 AI Trust Maturity Survey, ~500 organizations, Dec. 2025–Jan. 2026) — security and risk cited by ~2/3 of respondents as the top barrier to scaling agentic AI; only ~30% of organizations reach governance maturity adequate for autonomous systems.

11

Company research record: cross-architecture validation across 40 model architectures (410M–405B parameters), internally reported, not yet independently verified — see Validation section. Proprioceptive AI is funding up to $50K toward completion of agreed scope investor-directed technical due diligence.

13

Iregbulem, N. (Lightspeed Venture Partners), "Seed Valuations Aren't Valuations," public commentary, 2025–26.

15

Anthropic, "A Global Workspace in Language Models," Anthropic research publication, 2026.

17

Wurgaft, D., et al. (Goodfire), "Manifold Steering," 2026.

2

Gartner, "AI Spending Current Outlook, Worldwide, 2Q26," July 24, 2026 — enterprise AI spending forecast of $407B in 2026, up 34.8% year over year.

4

Magesh, V., et al., "Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools," Journal of Empirical Legal Studies 22, no. 2 (2025): 216–242.

6

Moffatt v. Air Canada, British Columbia Civil Resolution Tribunal, February 14, 2024 — company held liable for chatbot misstatement; “separate legal entity” defense rejected.

8

"The Range Shrinks, the Threat Remains: Re-evaluating LLM Hallucinations on the 2026 Frontier-Model Cohort," arXiv:2605.17062 — general-purpose frontier models measured at 4.6–6.1% hallucination rate on a ~40,000-prompt benchmark; open-ended generation and multi-step agent tasks measured substantially higher (15–40%) in separate industry benchmarking (Deepchecks, 2026).

10

McKinsey & Company, "The State of AI" (2025) — 88% of organizations use AI in at least one business function; PwC 2026 CEO Survey — only 12% of CEOs report both revenue gain and cost reduction from AI to date.

12

Company patent record: management has organized the working estate into 44 invention families and currently prioritizes approximately 17 leading prosecution targets under active review with outside patent counsel. Provisional filings; patent counsel confirming legal family structure and prosecution priority.

14

Crunchbase; company funding announcements (BusinessWire, TechCrunch, VentureBeat) for Goodfire, LMArena, Virtue AI, and HiddenLayer funding history, as cited.

16

Huang, H., LeCun, Y., Balestriero, R., "Semantic Tube Prediction," 2026.

18

Google Research, "ScientistOne / Science One Framework: Verifiable Autonomous Research via Chain-of-Evidence," 2026.

16

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

LEGAL

Disclaimer

For Institutional Investors Only; Not for Retail Investors

17

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

This preliminary presentation (“Presentation”) has been prepared by the issuer and is being delivered to the recipient (the “Recipient”) on behalf of the issuer (the “Issuer”) by Castle Placement, LLC (“Castle”), our placement agent. This Presentation and the information contained herein is CONFIDENTIAL and shall be kept confidential. By accepting this Presentation, the Recipient agrees it will not copy or distribute this Presentation, in whole or in part, at any time without Issuer's prior consent. Delivery of this presentation to anyone other than the Recipient is unauthorized and any reproduction of this preliminary Presentation, in whole or in part, or any attempt to divulge its contents, in whole or in part, without the prior written consent of the Issuer is prohibited. This Presentation is intended only for Qualified Institutional Investors and individual Accredited Investors as defined by all state and federal securities laws, the Securities Act of 1933 and Rule 501 of Regulation D. Castle does not provide legal, investment, accounting, regulatory or tax advice. As such, it should not be used as a substitute for consultation with professional legal, investment, accounting, regulatory, tax, or other competent advisors. This Presentation has been prepared solely for the purpose of providing a preliminary introduction to the opportunity described in the Presentation to assist a limited number of parties in deciding whether they wish to proceed with a further review of this opportunity. Information contained herein has been obtained from management of the Issuer and other sources which are deemed reliable, but no representations or warranties are made by Issuer or the placement agent or any of their respective affiliates, employees or representatives as to the accuracy or completeness of such information or any other information (whether written or oral) made available in connection with any investigation of Issuer. Neither the placement agent nor Issuer has independently verified any of the information set forth in this Presentation. Only those particular representations and warranties which may be made in a definitive agreement when, as and if it is executed, and subject to such limitations and restriction as may be specified in the definitive agreement, will have any legal effect. This Presentation does not constitute an offer or solicitation to sell or purchase any securities. The information contained in this Presentation does not purport to contain all of the information that a prospective purchaser may desire. In all cases, interested parties should conduct their own investigation and analysis of Issuer and the information contained herein. The information contained in this Presentation is not to be used for any other purpose. Except as otherwise expressly indicated, this Presentation speaks as of the date hereof. Neither Issuer nor the placement agent undertakes any obligation to update the information contained herein, correct any inaccuracies that may become apparent or provide the recipient with access to any additional evaluation material. The delivery of this Presentation does not imply that there has been no change in Issuer’s affairs after the date hereof. This Presentation is not a prospectus, disclosure document or offering document, and does not constitute an offer or invitation to apply for securities under any law. In particular, this Presentation shall not form the basis of or be relied on in connection with any contract or commitment whatsoever. This Presentation includes certain forward-looking statements, estimates and projections provided by the Issuer with respect to its anticipated future performance. Such statements, estimates and projections reflect various assumptions and judgments by the Issuer concerning anticipated results, which assumptions are inherently uncertain and may or may not prove to be correct. No representations or warranties are made as to the accuracy of such statements, estimates or projections. Nothing contained herein should be relied on as a promise or representation as to Issuer's future performance. In addition to forward looking statements, this presentation contains information and statistics relating to Issuer’s industry and related industries. Such information and data was derived from third party and other sources that it believes to be reliable, but without independent verification. Financial information contained herein has been provided by the Issuer. Neither the Issuer nor Castle warrants or guarantees this information, nor makes any representations as to the accuracy of the information. Recipient must verify this information independently, and represents that it has not relied on the Issuer or Castle with respect thereto. Issuer and the placement agent and their respective affiliates, employees and representatives expressly disclaim any and all liability relating to or resulting from the use of this Presentation or any errors therein or omissions therefrom, or in any other written or oral communication transmitted or made available to interested parties. Issuer reserves the right to require the return of this Presentation at any time. Under no circumstances should the Issuer or any of its employees, agents, owners, officers, directors, or members be contacted directly. Issuer reserves the right to (1) negotiate with one or more prospective buyers at any time and to enter into a definitive agreement with respect to itself and the Issuer without prior notice to the recipient or other prospective purchasers, (2) terminate further participation in the investigation and proposal process by any party, and (3) modify the procedures relating to this process without assigning any reason therefor. This Presentation is not a research report under U.S. law, not a research department product and is intended only for sophisticated investors/clients. Some or all of this Presentation may have been prepared by, and is a product of, a group that may support underwriting, sales, trading or investment banking activities, and Issuer does not hold it out as impartial in relation to such activities. These sources and uses represent current objectives. Several factors could result in material changes to the actual amounts, distribution and structure of how capital is invested and deployed, including without limitation: changes to the capital structure, strategy modifications, evolving technology and competition, market, economic, political, and regulatory conditions, management team changes, variances between financial and operational projections and results, or other unanticipated factors.

18

Unless otherwise referenced, all statements, projections and opinions were provided by Company management team. Past performance is not indicative of future results, and there can be no assurance that similar results will be achieved. Historical data and a detailed financial model with assumptions and scenario analysis functionality are available. Target returns and financial projections are presented solely to show Company's objectives, should not be used as a primary basis to invest, and there is no assurance of future performance, or the adequacy of the methodology used. Private securities are speculative, illiquid, and carry a high degree of risk – including the loss of the entire investment. For use by institutional investors only; not for use by retail investors.

Industry

AI

Company Type

Corporation

Location

Delaware, US

Size

$6 million

Investment Type

SAFE

* Registered trademark and subjective slogan of Castle Placement®

Richard Luftig

(212) 418-1181 · rluftig@castleplacement.com

Ken Margolis

(212) 418-1188 · kmargolis@castleplacement.com

Axel Reijmers

(201) 747 4718 · areijmers@castleplacement.com

Castle Placement, LLC

New York | Atlanta | Baltimore | Boston | Charlotte | Columbus | Denver | Detroit | Houston | Los Angeles | Miami | Salt Lake City | St. Louis

Headquarters

1460 Broadway

New York, New York 10036

castleplacement.com