▼   PRIVATE BETA · Q3 2026 Now accepting waitlist enrollments

AI that earns the
right to be wrong

Most AI is built to give you an answer. HALO is built to stay silent until it can prove one. When the answer matters — your regulator, your courtroom, your reputation — confident is not the same as correct.

DSRPT HALO is enterprise AI for situations where being wrong has consequences. Built for the people who have to defend the answer afterwards.

High-stakes decisions
Confident guessing
Q3 2026
▼   The Shift

The AI you have was built to help. The AI you need is built to hold up.

One is great for drafting an email. The other is what you reach for when a finding is going to be reviewed by a regulator, contested by opposing counsel, or read by a journalist who is going to call the people you named.

Generic AI assistants
HALO is built to defend its answers
Always gives you an answer
Stays silent when the evidence isn't there
Confident-sounding output
Every finding ships with the proof behind it
"Just trust me"
Independent third-party audit, every year
Names people without warning them
Built-in right of reply for anyone named
Output you can't reproduce
Every conclusion is traceable to its source
§01 When You'd Reach For It

The kind of work where almost right is wrong.

A General Counsel

Before a regulatory filing.

"We need to know what's actually in our exposure here. If we miss something material and the SEC finds it later, that's our license."

HALO surfaces patterns across filings, transactions, and public records — and refuses to flag a pattern unless it can prove it's there. Either way, you get a defensible record of what was checked.

A Compliance Officer

Sanctions screening with consequences.

"False positives are expensive. False negatives are fatal. Our team can't manually review every alert — but we can't afford to miss one either."

HALO doesn't generate alerts; it earns them. Every flagged entity comes with the evidence chain, the cross-references, and the competing explanations that were considered and ruled out.

A Due-Diligence Team

Pre-acquisition target review.

"We have six weeks to decide on a $400M acquisition. The target's books look clean, but something feels off. We can't articulate what."

HALO maps connections that no single document reveals — between beneficial owners, legal entities, regulatory filings, and public incidents — and shows you the structure the deal team needs to confront before signing.

An Internal Investigation

Before the news cycle starts.

"Allegations have been raised about a senior executive. The board wants the truth before the press does. We have one chance to get this right."

HALO produces a finding that can survive cross-examination — including a built-in right-of-reply for the named subject and a methodology trail your outside counsel can defend.

A Risk Officer

Mapping foreign-influence exposure.

"Three of our portfolio companies have ownership chains that pass through jurisdictions we can't audit. We need to know what we're holding."

HALO traces ownership through opaque structures, surfaces the beneficial-owner exposure, and names the gaps it cannot close — so your risk register reflects reality, not wishful thinking.

An Investigations Editor

Publishing on a powerful subject.

"We have the story. We need to make sure the lawyers can stand behind every sentence — and that the subject has had a fair chance to respond."

HALO is the engine behind DSRPT's own accountability journalism. It enforces the right-of-reply window, anchors every claim to a primary source, and produces an editorial record that withstands legal challenge.

§02 How It Works

Three steps. No black box.

01

You give it a question that matters.

A regulatory exposure check. A counterparty review. A dossier on a pattern of behavior. Anything where "approximately right" isn't good enough.

02

It investigates — and refuses to fabricate.

HALO pulls from primary sources, cross-references across domains, tests its own hypotheses, and surfaces what the evidence supports. If the evidence isn't there, it tells you that — clearly, instead of guessing.

03

You get a finding you can defend.

Every conclusion ships with its proof, its methodology, the competing explanations that were ruled out, and a notice to anyone named giving them a chance to respond. It's not a chat — it's a record.

§03 What It Solves

The pain points you've been working around.

01

Your AI invents things and you can't tell when.

HALO is built so that fabrication is structurally impossible — not "minimized" or "reduced." Every claim is anchored to a primary source the system can show you. If it can't show you, it doesn't say it.

02

You can't show your auditor how the AI decided.

HALO produces a methodology trail with every output. Your auditors get a record they can verify — not a vague "the model said so." Independent third-party audit is built into the contract, not optional.

03

Outputs that age into liability.

A finding that was correct in March might be wrong by August. HALO monitors its own published findings for new evidence that contradicts them. When a finding ages out, you hear about it before your adversary does.

04

You name people in reports — and pray.

When HALO names a person or entity, it automatically opens a right-of-reply window. The subject has a fair chance to respond. The response is published with the finding. This is what defamation lawyers call "due diligence."

05

You're worried about being trapped with a vendor.

HALO's underlying specification is openly published and irrevocable. If we go away tomorrow, your audit trail is still verifiable, your data is exportable, and any auditor can confirm the integrity of every record.

06

You can't customize away the safeguards.

You can scope HALO to your domain. You cannot weaken the safeguards. Right of reply, evidence thresholds, falsifiability checks, audit cadence — these are locked. We tell you this clearly because mismatched expectations are how procurement falls apart.

§04 What's In The Box

Everything you need to defend the answer.

01

Refuses to make things up.

If the evidence isn't there, HALO doesn't speak. Most runs return less than you'd expect — and that's the system working as designed.

02

Cross-domain pattern detection.

Connects evidence across financial filings, public records, news archives, regulatory disclosures, and more — surfacing patterns no single source reveals.

03

Built-in right of reply.

When HALO names someone, it tells them. Their response is part of the published record. The lawyers love this; the named subjects respect it.

04

Tamper-evident records.

Every finding, every revision, every decision is sealed cryptographically. Anyone can verify a record is original — even years later, even after we're gone.

05

Independent annual audit.

Within 90 days of going live, an outside firm — never one we have a relationship with — re-runs our audit protocols. The findings are published.

06

Watches its own work age.

Three monitoring systems track every published finding for new evidence, missed knowledge, and methodology drift. You hear about issues before opposing counsel does.

07

Cryptographic data deletion.

When you ask us to delete your data, we don't just remove pointers — we destroy the keys. The data becomes mathematically unrecoverable. Faster than scrubbing. Verifiable.

08

Open specification.

The standard HALO operates against is openly published and irrevocably licensed. You're not buying lock-in. You're buying a service that meets a standard you can verify.

09

Discloses our relationships.

If HALO produces a finding about a customer, vendor, or partner of ours, that relationship is automatically disclosed in the output. Conflicts surface before opposing counsel finds them.

10

Multi-jurisdiction posture.

We operate where source-protection laws are strong. We decline jurisdictions where they're weak. Our shield-law map is transparent.

11

Clear kill switches.

Eight levels of stop, from pausing a single agent to halting the whole system. If something goes wrong, we don't have to debate what to do — the procedure is written.

12

Subpoena-ready architecture.

What we'd produce in response to a court order — and what stays sealed for in-camera review — is documented up front. Not invented under pressure.

§05 What Customers Will Tell Their Boards

The answer matters. The defense of the answer matters more.

We haven't shipped yet. Below are the outcomes we're designed to produce — drawn from the conversations that brought HALO into existence, with the kinds of customers who reach for tools like this when nothing else has worked.

The Compliance Lead Mid-cap financial services
The auditor asked how we got there. For the first time, we had a one-page answer they could check.
What changed The team spent 11 months building an explainability deck for the prior system. With HALO, the methodology trail is generated automatically with every finding. They retired the deck.
The General Counsel Global manufacturing group
We were going to publish. Then HALO surfaced the counter-evidence we missed. We didn't publish — and that decision saved us a defamation suit.
What changed The team learned that the absence of an output is itself an output. A "we are not confident enough to say this" record is sometimes more valuable than a finding.
The Investigations Director Public-interest newsroom
Six weeks of reporting condensed into a record the legal review could actually defend. The subject responded. We published the response. We won the suit.
What changed Right-of-reply, methodology trail, and primary-source anchoring stopped being three separate workflows. They became the default output of every investigation.
§06 The Receipts

Promises we put in writing.

90D
Maximum window from your launch to an independent third-party audit. Contractually mandated.
0
Records ever modified after the fact. The audit trail can't be rewritten — by anyone.
14D
Default right-of-reply window for any named subject. Built into the system, not bolted on.
CC BY 4.0
License on the underlying specification. Irrevocable. If we vanish, your audit trail still works.
HALO
§07 Waitlist · Beta · Procurement

Tell us what you cannot afford to be wrong about.

Private beta opens Q3 2026 to qualified enterprise customers. Limited cohort. We will route inquiries to the right contact within five business days — no marketing automation chasing you in between.

By enrolling you accept the privacy notice. We will only contact you about HALO.
Received. Thank you.
We will route within 5 business days.
§08 Questions People Ask First

The hard questions.

Is this a chatbot?

+
No. HALO doesn't chat. It doesn't summarize meetings or write emails. It produces findings — the kind of answer a regulator, an opposing counsel, or a journalist might examine line by line. Most days, you'd use a tool like ChatGPT for the helpful, drafty work, and HALO for the work that has to hold up.

What if HALO doesn't find anything?

+
That's the system working. HALO refuses to manufacture a finding just because you asked for one. If the evidence isn't there, you'll get a clear "we are not confident enough to say this" — with a record of everything that was checked and why it didn't reach the bar. That record is, surprisingly often, more valuable than a finding would have been.

Who is HALO actually for?

+
People who have to defend the answer afterwards. General counsel. Compliance officers. Investigators. Risk and audit teams. Due-diligence firms. Investigative editors. Any function where AI output could end up in front of a regulator, in a courtroom, or on the front page.

How does HALO compare to general-purpose AI like ChatGPT or Claude?

+
They solve different problems. General-purpose AI is built to be helpful. HALO is built to be defensible. You'd never use HALO to draft a press release; you'd never use a general assistant to produce a finding for the SEC. We're not competing — we're a different category.

What if you go out of business?

+
The technical specification HALO operates against is published openly and licensed irrevocably. If we vanish tomorrow, your audit trail is still verifiable, your data is still yours and exportable, and any third party can independently confirm the integrity of every record. You're not buying lock-in. You're buying a service.

Can we customize the rules?

+
You can scope HALO to your domain — define what you're investigating, what counts as a primary source for you, what your kill criteria are. You cannot weaken the safeguards. Right-of-reply, evidence thresholds, audit requirements — those are locked. If you need looser thresholds than that, HALO is the wrong tool. We say this clearly because procurement falls apart on mismatched expectations.

How do we evaluate it before signing?

+
The technical specification and conformance harness are public — your engineers can run them today against your own systems, no DSRPT relationship required. For deeper evaluation, qualified enterprises can request a scoped paid pilot — outcome-linked pricing, partial credit if we don't deliver. Join the waitlist above and we'll route to the right starting point.

Where are you based?

+
DSRPT is based in West Palm Beach, Florida. We host data and operate in jurisdictions with strong source-protection and journalist-shield laws. Where those protections are weak, we don't operate. Our full jurisdiction map is available under MSA execution.