FAQs

R3nquist frequently asked questions

About R3nquist

R3nquist is an agentic technology company. We unite data and intelligence in one governed platform to deliver defensible outcomes in litigation and cyber incident response.

Our agents do the volume work, and every decision they make is grounded in the source documents, logged and reproducible. Our people are the other half of the model. R3nquist is built by people who have run discovery and breach response at scale, and who have built and governed autonomous agents. They set the protocols the agents work to, handle the exceptions and stand behind every result. Agent-dominant, not human-absent.

The name works on two levels: a standard and an architecture.

The standard comes from Chief Justice William Rehnquist, known for intellectual rigor, precision and clarity in legal reasoning. For him, thoroughness and defensibility were not aspirations; they were the baseline. Those are the principles we bring to every matter.

The R3 reflects how our platform works: reasoning, retrieval and response, the core building blocks of agentic AI, brought together in one coordinated system.

R3nquist was founded by Dan Reed, Founder and Chief Executive Officer, who previously founded UnitedLex, and Julian Box, Founder and Chief Innovation Officer, who co-founded Virtustream.

R3nquist works with law firms, in-house legal teams, and insurers and the panel firms they appoint.

Our clients are in the United States, Canada and the United Kingdom. Matters often involve data held elsewhere, such as in the European Union; where they do, we agree at the start where the data is held and how it may lawfully be moved.

Services

R3nquist offers two services: Cyber Incident Response and eDiscovery. Both run on the same governed platform and are delivered under the same three commitments: extraordinary value, days not months, and governance designed for regulatory change.

After a cyber incident, R3nquist establishes exactly whose personal data was exposed. Our agents do that work under defined supervision, inside a closed platform, from data mining to the final list.

The result is a definitive list of affected individuals that you can stand behind, with a recorded basis for every decision to include or exclude someone.

R3nquist eDiscovery applies the same engine to litigation discovery. Our agents carry out the review work under human governance, and every coding call is grounded in the source document, logged and reproducible.

Your lawyers set the protocol, review the sampling and sign off the output. Results are validated with the measures courts already accept: recall, precision and elusion on a control set.

Because the two answer different questions. A forensic (DFIR) firm establishes how the attacker got in and what happened on your systems. R3nquist establishes what sensitive data was affected: which individuals it belongs to, what kinds of personal information about each of them were exposed, and therefore who must be notified, and under which laws.

That data review is high-volume work, and it is the work our agents do, with measured accuracy and a recorded basis for every notification decision. We work alongside your forensic firm, and your lawyers govern both.

No. The aim is a better outcome, not a migration. R3nquist can work alongside the tools you already use, and should be judged on what it delivers: a smaller review population, fewer review hours, faster completion and lower cost.

A proof of concept measures those outcomes without any change to your platform or production workflows.

Most AI in legal technology assists a person who is still running the review: it suggests coding, summarizes documents or predicts relevance, while people run the workflow, the queues and the quality control.

R3nquist's agents carry out the workflow itself, under human governance. It is not AI-assisted; it is AI-executed. Every decision comes with its own record of which agent made it, or which person changed it, why, and on what evidence, so the whole process can be explained, reproduced and put in front of a court or regulator.

Pricing and delivery

R3nquist prices the result, not the effort. You get one fixed price, set at intake. It does not move when the data proves larger, denser or harder than anyone expected. The only variable is how quickly you need the answer, so you know the number on day one.

R3nquist compresses production timelines from months into weeks or days, and the delivery date is set before the work starts. Your deadline is fixed. Ours is too.

No. With R3nquist, speed only matters if the result is defensible, so every time saving comes with the validation behind it. We measure accuracy on each matter with recall and precision rather than assuming it, and every decision is logged so it can be checked.

In incident response this matters most: notifying too many people or too few both create risk, so the speed is in service of a measured, defensible list.

Defensibility and validation

R3nquist grounds every decision in the source document, and logs it so it can be reproduced. We validate the output with the measures courts already accept: recall, precision and elusion on a control set. Your lawyers set the protocol and sign off the output.

Hallucination, where AI invents facts, comes from open-ended generative chat, where a model writes answers without being tied to a source. Discovery is a different kind of task: classifying, identifying, prioritizing and escalating documents.

R3nquist's agents reason over your actual documents and cite them. They retrieve; they do not invent. Every decision traces back to the underlying document in one click, and we validate and measure decisions rather than taking them on trust.

Yes. R3nquist calls it chain of custody for reasoning. For every material decision you can see which agent acted, why it acted, what evidence it considered in the source documents, how confident it was, whether and how the decision was escalated, and where a person changed the result.

That is more than most traditional review can show once reviewer numbers grow. It also means the process is not a black box: every step can be audited, traced and reproduced.

Agent-dominant, not human-absent. R3nquist's agents carry out repeatable, high-volume work under defined human supervision, and lawyers govern the legal decisions. If an agent is unsure, it raises its hand: our exception-handling protocol stops the decision and brings in an experienced human reviewer.

People sit at the checkpoints that matter: your lawyers set the protocol, review the sampling and sign off quality control. Experienced reviewers move from working through queues to overseeing and checking the process.

No method guarantees zero privilege misses: not human review, and not technology-assisted review. What matters is that privilege is handled as a measured workflow rather than an assumed outcome.

R3nquist measures privilege identification, escalates uncertain calls to a human reviewer and applies secondary quality control. Every privilege call is traceable and reviewable.

Getting the notification population wrong in either direction creates regulatory and legal exposure, so R3nquist treats accuracy as something to measure, not assume.

Our agents identify sensitive data in the source material, with recall and precision measured on that identification. The same individual often appears across many sources under slightly different details; we resolve those duplicates so each person is counted once. Every decision to include or exclude someone is logged with its basis, giving your lawyers a regulator-ready record.

A clean process with a clearly documented purpose also supports your lawyers' position on privilege. Courts have ordered breach forensic reports to be produced where the process and its purpose were poorly documented. The privilege call belongs to your lawyers; our role is the disciplined, documented process behind it.

That depends on the court and the matter, and is for your lawyers to decide. Where a court requires disclosure of AI use, R3nquist's documented and validated process is an asset: it is the same reasonable-process account parties already give for technology-assisted review.

We provide methodology and validation documentation that gives your lawyers a clear, sufficient answer to disclosure and meet-and-confer questions.

No method, human or machine, removes all error. With R3nquist, responsibility is handled the way it always has been in discovery: a reasonable process, appropriate court protections where your lawyers seek them, and clear contractual terms.

Your agreement with us sets out liability and indemnities explicitly. And because every decision is evidenced and reproducible, the process is easier to defend than an undocumented review.

R3nquist was built for the hardest domains on purpose: litigation and cyber incidents, where data is largest, most varied and most urgent. We measure validation on every matter, so performance on difficult data is surfaced and reported rather than hidden.

If a data type or language is outside scope for your matter, we tell you at the start rather than letting quality drop quietly. The best test is your own hardest data, scored on metrics agreed in advance.

Yes, and we recommend it. The most credible evidence of what R3nquist can do comes from your own matters, custodians, documents and review protocols, not from vendor benchmarks.

A proof of concept is scoped to your data, with success metrics agreed up front so the results are yours. It can run in parallel with your current process, with no change to your production workflows.

Security

R3nquist's information security management system is certified to ISO/IEC 27001:2022, and we have completed a SOC 2 Type II examination, in which an independent auditor tested the design and operating effectiveness of our security controls over a period of time. Copies of the certificate and report are available to customers and prospective customers on request, on confidential terms. ISO/IEC 27701 (privacy) and ISO/IEC 42001 (AI management) are on our certification roadmap.

How we protect, store and use your data is covered in the next sections.

Your data

You do. We process your data on your behalf and only on your documented instructions. Apart from a license to use the synthesized data described below, nothing in our agreement transfers any right in your data to us.

We use your data only to deliver the work in your statement of work. Before we return or destroy it at the end of an engagement, we also create synthesized data from it to improve our tools, unless you ask us not to. We never sell or license your data, never combine it with another client's data, never use it in another client's work, and never use it for marketing or to profile anyone beyond the work you have asked for.

The only third parties who handle your data are approved sub-processors, listed in your agreement. Under our standard terms you get at least 30 business days' notice before we add or replace one, and the right to object.

Only the people named to your engagement, which can include vetted document review and data collection partners engaged as approved sub-processors and listed in your agreement. All of them are screened, bound by written confidentiality obligations and trained in handling matter data. Access uses multi-factor authentication, is logged, and is reviewed at least quarterly under our standard terms.

Each matter runs in its own isolated tenancy, separate from every other client's data. Your data is encrypted at rest and in transit, and is never held on personal devices, personal cloud services or removable media, or in development, test or staging environments. Our independent security certifications are covered above.

Only in the locations agreed for your engagement and recorded in your agreement, along with the locations from which remote access is allowed.

We will not move your data outside those locations without your prior written consent and a lawful transfer mechanism, such as an adequacy decision, standard contractual clauses, the UK International Data Transfer Agreement or Addendum, or binding corporate rules, backed by a transfer impact assessment where the law requires one. Remote access from anywhere outside the agreed locations counts as a transfer, and the same rules apply.

AI and model learning

Not on your actual records. None of them is ever used to train our systems: not a copy of them, not a sample of them, not any part of them. Your real data never enters the separate, secure environment where our tools are trained.

Two things can happen, both explained below. During your engagement, we may adjust our tools to work well on your material; those adjustments are used only for you and discarded when the engagement ends. And unless you ask us not to, we create synthesized data at the end of the engagement: new, made-up records based on the overall shape of your material, which we use to improve our tools.

We don't take your data and simply remove the names. We create new data from scratch.

When an engagement ends, and before we destroy our copy of your data, a specialist company studies the overall shape of the material: what types of information it contains and how they connect. It is an approved sub-processor, listed in your agreement, and works on your data only inside your engagement environment. From that shape it creates a completely new set of made-up records, called synthesized data, in which every name, number and detail is invented.

Anonymization, the usual alternative, keeps a copy of your data with the names removed, and small clues in that copy can be enough to identify someone. Synthesized data leaves no copy of your data in our training material, so the risk of anything leaking or being pieced back together is far lower, and the checks below are designed to catch what remains.

Practicing on synthesized data makes our tools faster and more consistent, including on any further work for you. Synthesized data may be combined with synthesized data from other sources, is held at our highest confidentiality level, and is kept for no more than five years under our standard terms.

Every set of synthesized data goes through four checks:

  • Before the new records are created, we list every personal detail in your material (every name, email address, phone number or anything else that could identify a person) plus every mention of your organization's name. The list is treated as your data: it stays inside your engagement environment, only your engagement team can access it, and it never enters the environment where our tools are trained.
  • Once the new records are created, we check them against that list, including close variants and partial matches such as misspellings and abbreviations. Any record containing one of those details is destroyed and never used.
  • Our Data Protection Lead, or a colleague they nominate who was not involved in creating the records, confirms that nothing in them identifies a real person or your organization. We record who did the check and when.
  • The list itself is deleted with the rest of your data when the engagement ends.

Before we release an updated model, we test it to reduce the risk that it reproduces content from its training data. Data under a legal hold is never used. You can tell us about any material that must not be used, or ask us not to create synthesized data at all; if you opt out, your agreement is drawn up on that basis.

Never. We do not put your data, or any excerpt, summary, paraphrase, screenshot or output from it, into any public AI service. All AI processing of your data happens inside the environment approved for your engagement.

Any third-party AI model provider we use is an approved sub-processor, contractually barred from using your data to train or improve its own models, with enforceable limits on how long it can keep it. Our AI agents work within defined permissions and task scope, and their actions and data access are logged. A competent member of the engagement team reviews AI output before it is relied on in anything we deliver to you.

Only for your engagement. We may adjust our models during an engagement so they work well on your material. Those adjustments are used only for your engagement, are discarded when it ends, and are never kept or used in work for anyone else. Your data is never used to improve the models we use across our services.

If your statement of work includes training a model just for you, that model is used only for you, held in your engagement environment and treated as your data for security, return and destruction. Who owns it is set out in your statement of work; if it says nothing, the model is destroyed when the engagement ends.

Incidents and the end of an engagement

You choose: we return it to you or destroy it, within 30 business days under our standard terms. If you haven't chosen by then, we remind you, and if we still don't hear within a further 14 business days, we destroy it. Destruction makes it unrecoverable, using methods consistent with NIST SP 800-88. Copies in backups are purged within the backup cycle or 90 days, whichever is sooner; until then they remain encrypted and cannot be accessed for any other purpose. Within ten business days of destruction, we give you a certificate of destruction signed by an authorized person.

We keep nothing afterwards except synthesized data (unless you opt out) and our audit records of what was done.

If you expect litigation, an audit or a regulatory investigation, you may need the data preserved. This is called a legal hold. Tell us in writing and, within two business days under our standard terms, we stop all deletion of that data, including automated deletion, and confirm it in writing. We keep it fully protected until you release the hold, and we never use it to create synthesized data unless you tell us in writing that we may. If the law requires us to keep any data for longer, we tell you what we are keeping and why, and destroy it when the obligation ends.

Under our standard terms, we notify you without undue delay, and within 24 hours of becoming aware, of any actual or suspected security incident affecting your data. We tell you what we know at that point, and as our investigation establishes them, what happened, what data and individuals are affected, the likely consequences and what we are doing about it. We keep you updated throughout. We cooperate with your investigation, preserve evidence and act to contain the incident. We will not notify a regulator or any affected individual on your behalf without your agreement, unless the law requires us to.

If a court, regulator or law enforcement body asks us for your data, we check that the request is genuine and lawful, tell you promptly (unless the law forbids it) so you can challenge or limit it, disclose only the minimum needed, and keep a record of what was disclosed. If an individual asks to exercise their data rights over information in your data, we pass the request straight to you and do not respond to it ourselves, beyond acknowledging it, unless you instruct us to.