For banks, insurers, hospitals, clinics, law firms and notaries. AI that works on your confidential documents without them ever leaving your computer.
You hold exactly the documents AI could help with most, and you can least send them to a cloud AI service. Antloom runs the AI on your own computers. No document leaves the organization.
Antloom shows what it found and where. Your people decide.
The problem Antloom solves.
Banks, insurers, hospitals, clinics, law firms and notaries hold exactly the documents AI could help with most. They are also the organizations that can least send those documents to a cloud AI service: professional secrecy, banking secrecy, medical confidentiality, the GDPR and their clients’ trust all stand in the way.
So many of them either go without AI or take a risk with their clients’ data.
With the AI on the organization’s own computers:
Nothing to vet or trust
There is no cloud provider to check, contract with or rely on.
Nothing uploaded
Documents are read where they are stored.
No internet needed
The work runs entirely offline.
Each application solves a business problem.
Document Search
Find anything your organization ever wrote, and get answers from it.
Your people get answers from your own documents and mail, with the passages each answer rests on, so they can check an answer before they act on it. That knowledge is scattered across shared folders, old contracts, reports, scanned PDFs and years of mail, and finding the clause, the email or the version that settled a question takes time. Most organizations cannot give confidential documents to a general AI chatbot, and many chatbots answer without showing where an answer came from.
Ask in plain words.
People ask a question and get the answer from their own documents and mail. The search matches meaning as well as exact words.
Every answer shows its sources.
Each one cites the exact passages it rests on, and every claim is checked against those passages before it is shown.
It reads everything.
Word, PDFs (including scanned ones), spreadsheets, presentations and mail.
Private stays private.
Confidential folders can be marked Private.
It fits the hardware.
It runs on the organization’s existing computers and adapts to each machine, so people can choose between faster and more thorough answers.
Document Checker
Your best reviewer’s eye on every document.
Every new draft comes back with comments based on the corrections your own experts made before. Your experts then review a draft that already carries those comments, and juniors see their seniors’ standards on every draft. Legal, notarial, financial and medical documents are checked by experienced experts whose time is scarce, and standards often differ from one reviewer to the next. General AI checkers do not know your organization’s standards, and most organizations cannot give them these documents.
It learns your experts’ standards from their own corrections.
Give it documents as drafted, beside the versions your experts corrected. It finds every correction, learns which kinds of mistakes your organization catches, and turns them into checks.
Every new document is checked automatically.
It comes back with comments on the exact words they are about, in Word, OpenDocument or a findings report. The original is never touched.
It reports its hits and false alarms.
For example: “caught 31 of the 40 corrections, raised 6 false alarms”, measured on documents it has never learned from.
It changes a check only after testing it.
It refines its own checks and keeps a change only when the test documents prove it loses nothing and gains something. Your experts’ right-or-wrong verdicts teach it further.
It handles the hard cases.
- Long documents are checked article by article.
- A kind of document that keeps tripping it up gets a second, specialized checker.
- A document unlike anything it has learned from is flagged as such.
Learning runs when computers are idle.
It stops the moment someone starts using the computer.
Compliance
A family of apps for EU regulation.
The compliance apps check your documents and code against EU law each time you run them, and keep the evidence ready for when an auditor or regulator asks. For financial, medical, legal and software-driven organizations, EU regulation is a growing stack of laws, and each of them changes:
privacy (the GDPR);
artificial intelligence (the EU AI Act);
financial resilience (DORA);
cybersecurity (NIS2 and the Cyber Resilience Act);
money laundering, payments, health data, medical software, and more.
Each law reaches into the policy documents, the processes and the software itself. In many organizations this is checked by hand, often once a year. Sending documents or code to a cloud tool for that check adds an outside company you must contract with and monitor.
Antloom’s compliance apps run on the organization’s own computers.
EU Compliance: see where you stand, documents and code alike
Documents and code together.
It reads the organization’s own documents (privacy notices, records of processing, processor contracts, AI system documentation, resilience and security policies, incident procedures) and its software’s source code (Java, Python, JavaScript and TypeScript, C#). It points out where they look short of the law.
The law’s own words beside every finding.
The exact article and the official text stand next to the organization’s own passage or lines of code.
Real problems in code, down to the line.
Examples:
- personal data flowing into logs;
- passwords in source;
- encryption switched off;
- AI components in use;
- cookies set before consent;
- vulnerable dependencies;
- a software bill of materials that doesn’t match reality;
- data stored outside the EU.
Missing documents count too.
A document the law requires but the organization doesn’t have is shown as clearly as a flawed one.
It keeps up with the law.
It follows the EU’s own official publications. New and amended law reaches your computers as an update, and it re-checks what a change affects. National law is included, starting with the Netherlands.
Tested on real cases.
The checks are tested against real cases, including documents that data protection authorities actually found deficient.
The compliance officer decides.
Every finding is confirmed or rejected by the organization’s own people, and those verdicts sharpen the checks. Every scan keeps an exact record of what was read, checked and found, against which version of the law, ready for auditors and regulators.
The coverage grows field by field.
The GDPR, the EU AI Act and DORA lead. NIS2 follows, with its cybersecurity duties and incident reporting deadlines for essential and important organizations, then the Cyber Resilience Act, the cookie rules, anti-money-laundering, payment services, consumer credit, crypto-assets, health data and medical device software.
The DPIA tool: data protection impact assessments, done properly
It finds where an assessment is needed.
It identifies the processing in the organization’s documents and systems that calls for a data protection impact assessment.
It checks each assessment is complete.
An assessment must cover what the GDPR requires: the description of the processing, its necessity and proportionality, the risks to people, and the measures that address them.
It keeps the evidence together.
It shows what is missing or thin in each assessment, beside the law’s own words.
For organizations with many assessments.
This is where much of the day-to-day privacy work sits.
Financial Reconciliation
Find out why invoices, payments and bank statements don’t match.
Every difference between invoices, payments and bank statements comes with the rows that explain it, so fewer hours go into matching rows by hand. These records sit in different systems and rarely match by themselves. Many finance teams match them in spreadsheets, and when a payout is lower than expected, someone goes through all three to find out why. A cloud tool would need the invoices, contracts and payment files uploaded.
It matches invoices to payments.
Each invoice is matched to its payment and to the line on the bank statement or payout export.
It lists the differences.
Missing, partial and unexpected payments are listed, each with the rows it is about.
You can ask why.
Ask why a payout is lower than expected. The answer comes from your own records, with the rows and documents it is based on.
For banks, insurers, payment and e-money institutions.
The rules they work under
Under DORA, financial institutions keep a register of information on all their contracts for technology services from outside technology suppliers (Article 28(3) of Regulation (EU) 2022/2554). Under the GDPR, every outside company that processes personal data for them must work under a contract that sets out how (Article 28(3) of Regulation (EU) 2016/679). Antloom runs on the institution’s own computers, so no cloud AI provider is added to the register and no client file is sent to one.
What Antloom does there today
Document Search on contracts, policies and client files.
Staff ask a question in plain words and get the answer from the institution’s own documents, with the passages it is based on.
Document Checker on reports and filings.
New reports and filings are checked against the standards of the institution’s own experts. Each comment points at the words it is about.
Compliance for DORA, the GDPR and the EU AI Act.
Each finding is shown next to the text of the law. The compliance officer makes the call.
Financial Reconciliation on invoices, payments and bank statements.
Invoices are matched to payments and bank statements, and every payment that does not match is listed with its rows.
In the pipeline
Agents for payment and banking systems.
Agents that work with payment and banking systems for the institution.
More in the pipeline.
More compliance apps and fields.
As listed above, starting with NIS2.
More apps on the same platform.
Each runs on the organization’s own computers.
Antlooms that work together.
An organization’s computers will be able to share their AI processing with each other, safely and under the organization’s own key. A busy laptop can use a colleague’s idle machine.
One technology foundation. Two strictly separate editions.
For individuals
Launching soon
Find almost anything on your computer.
Use AI powered search to find and understand your own files. The application runs locally. Your documents stay on your device.
Includes
More to follow
- Programming
- Writing books and articles
Free at home.
For organizations
Launching soon
Use advanced AI inside your own environment.
Run AI applications on computers you control, including computers disconnected from the internet. Each run can be recorded and replayed.
Includes
The two editions remain completely separate. A business deployment never connects to the consumer computing network.
The promise
Document Search
Every answer cites the exact passages it rests on.
Document Checker
Every comment sits on the exact words it is about, and its accuracy is measured on documents it has never learned from.
Compliance
Every finding sits beside the law’s own words, and your own people confirm or reject it.
The technology underneath: Antloom core
Each result is cross-checked before you see it. Several AI models that run on your own computers read the same evidence, each with its own task. They compare their conclusions and check every disagreement against the evidence.
Who builds Antloom.
Tijn van der Zant
PhD in artificial intelligence, University of Groningen. Co-founder of RoboCup@Home, the worldwide competition for robots in homes. Building AI since 1996. LinkedIn
Dirk Meuleman
Chairman and chief executive officer of Phenix Capital Group, which provides research, ratings and advice on impact investment funds to pension funds, insurers and banks. Earlier, he managed private equity and infrastructure portfolios at MN and at Shell. He has sat on the advisory boards of more than twenty private equity funds. LinkedIn
Email us
info@antloom.com