Acceptable use policy

1.Purpose and scope

This policy sets out how DocxIntel may and may not be used. DocxIntel is a product of BizfyLabs FZC LLC, and this policy forms part of the terms and conditions at docxintel.com/terms-conditions and of every licence agreement BizfyLabs FZC LLC enters into.

It exists for two reasons. The first is commercial: a licence carries limits, and the deployment bundle contains third-party model weights whose own licences must be respected. The second is that a document intelligence system reads identity documents, medical records, claims and contracts, and a tool that reads that material can be misused. We would rather state the boundary plainly than discover a disagreement about it later.

Because DocxIntel runs inside your infrastructure, we have no technical ability to monitor how you use it and no wish to acquire one. There is no telemetry channel and no usage reporting. This policy is therefore a contractual commitment you give, not a control we enforce from the outside.

• It covers the licensed software, the model weights and other components shipped in the deployment bundle, and any evaluation or Proof of Value environment.
• It covers the docxintel.com website and the documentation portal.
• It sits alongside, and does not replace, the licence restrictions in the terms and conditions and the licence terms of each third-party component.

2.Who this applies to

This policy applies to every person and organisation that receives, installs, operates, administers or uses DocxIntel, whether under a full licence, an evaluation licence or a Proof of Value statement of work.

• The licensee named in the order form, and each affiliate named there as permitted to use the deployment.
• Your employees, contractors, consultants and managed service providers who operate or administer the deployment on your behalf.
• Any system integrator, reseller or delivery partner acting for you, whose compliance you are responsible for.
• Any end user inside your organisation who submits documents to a deployment or consumes its output.
• Visitors to the docxintel.com website and readers of the documentation.

You are responsible for making the people above aware of this policy and for their compliance with it. A breach by someone acting for you is treated as a breach by you.

3.Permitted use

Within the scope of your licence, and subject to the limits below, you may use DocxIntel freely and without metering. There is no page count to manage and no per-document charge, so the permitted envelope is generous by design.

• Processing your own documents, and documents you have been lawfully given the right to process, for your internal business purposes.
• Running the deployment in production, in staging and fully air-gapped, in the number of deployments recorded in your order form.
• Reprocessing your entire archive, as often as you like, including after a model or configuration update, at no incremental licence cost.
• Retaining, storing, indexing and further using the outputs the software produces from your documents, inside your own systems.
• Integrating the deployment with your own systems through its API, and building your own applications and workflows on top of that output.
• Configuring field schemas, taxonomies, confidence thresholds, prompts and routing rules to suit your document mix.
• Evaluating the software against your own documents and sharing the results internally, including with your auditors and your regulator.

4.Prohibited use

You must not use DocxIntel, or permit it to be used, for any of the following. This list is not exhaustive: anything unlawful under the law that applies to you is prohibited whether or not it appears here.

Processing documents you have no right to process

Submitting documents or personal data that you do not hold the necessary rights, consents or lawful basis to process, or that were obtained unlawfully — including material acquired through a data breach, theft, unauthorised access or deception.

Unlawful surveillance or profiling

Using the software to build or feed a system for unlawful monitoring of individuals, for covert tracking, for suppressing lawful political or journalistic activity, or for social scoring of individuals.

Unlawful discrimination

Using extraction, classification or generated output to treat individuals less favourably on a basis prohibited by applicable law, or to construct a proxy that achieves the same result indirectly.

Circumventing licence limits

Running more deployments than your order form permits, tampering with or bypassing a licence key or entitlement check, continuing to use the software after the licence term has ended, or sharing the deployment bundle with a third party outside your licence.

Extracting or redistributing model weights

Copying, exporting or extracting the model weights out of the deployment for use in another system, product or model; distilling, cloning or attempting to reconstruct them; or redistributing them in any form outside the permissions granted by the applicable component licence.

Reverse engineering

Decompiling, disassembling or otherwise attempting to derive the source code, architecture, training data or internal parameters of the software or the models, except where mandatory law expressly permits it.

Publishing benchmarks without consent

Publishing a benchmark, comparative test, accuracy claim or performance review of the software externally without our prior written consent. We do not withhold consent unreasonably where the methodology, document mix and version are disclosed and the result is reproducible — and we will correct our own published figures if you show us they are wrong.

Automated decisions without human review

Using extracted or generated output as the sole basis for a decision that produces a legal effect on an individual, or a similarly significant effect — such as refusing a claim, declining an application, terminating a benefit or reporting someone to an authority — without meaningful human review of that output.

Onward service provision

Using the deployment to provide a document-processing bureau, outsourcing or software-as-a-service offering to third parties, unless a written reseller or service-provider addendum permits it.

Harmful or deceptive output

Using the modification and redaction capabilities to forge a document, falsify a record, remove a legally required disclosure, or create material intended to deceive a court, a regulator, a counterparty or an individual.

5.Security and integrity

The deployment runs on your infrastructure, so its security posture is largely in your hands. These rules protect the integrity of the software, of the components inside it, and of the environments other customers run.

• Do not tamper with, modify or repackage the container images, model artefacts or signed manifests in the deployment bundle, other than through the supported configuration mechanisms.
• Do not remove, obscure or alter any copyright notice, attribution, licence file or product identification in the software or the bundle.
• Do not introduce malicious code into a deployment, or use a deployment as a staging point for an attack on another system.
• Do not attempt to probe, scan, penetrate or disrupt BizfyLabs infrastructure, the documentation portal or the website, other than under a written authorisation from us.
• Do not share licence keys, deployment bundles, private registry credentials or administrative credentials outside the people permitted by your licence.
• Keep your platform within the supported version matrix and apply the security updates we publish, so that a known and fixed defect does not remain exposed.
• Protect the outputs as carefully as the source documents. Extracted fields are frequently more sensitive than the page they came from, because they are structured and searchable.

6.Export control and sanctions

The software and the model weights shipped with it may be subject to export control and sanctions laws, including those of the United Arab Emirates and of jurisdictions applicable to the components we distribute. Because you receive the weights rather than calling a hosted service, these obligations travel with the bundle.

• Do not install, use or make the software available in a jurisdiction, or for an end use, prohibited by applicable export control or sanctions law.
• Do not provide access to a person or entity subject to applicable sanctions, or owned or controlled by one.
• Do not transfer, re-export or onward-supply the deployment bundle or the model weights to a restricted destination or party.
• Do not use the software in connection with weapons of mass destruction or any other prohibited end use.
• Obtain any licence or authorisation your own activities require, and tell us if a change in your circumstances affects your ability to give these confirmations.

7.Responsible use expectations

Document intelligence is probabilistic. Every value DocxIntel returns carries a page number, bounding-box coordinates and a confidence score precisely so that a human can check it, and the product is built on the assumption that someone will. These expectations are how that assumption is honoured.

• Set confidence thresholds deliberately and route low-confidence output to a reviewer rather than accepting it silently. A flagged 2% that reaches a human is the point of the design.
• Keep a human in the loop for any decision with a material effect on an individual, and make sure that reviewer has the authority and the time to change the outcome.
• Give reviewers the source region alongside the extracted value, so review is a genuine check against the original page rather than a rubber stamp on a number.
• Do not present generated answers as verified fact. Answers produced by the Ask capability carry citations back to the source passage — retain and surface those citations.
• Validate performance against your own document mix before going live, and re-validate after a model update rather than assuming parity across versions.
• Monitor for drift: track override rates and error patterns by document type, and investigate a rise rather than raising the threshold to hide it.
• Tell affected individuals, where the applicable law requires it, that automated processing is used in a decision that affects them, and provide the route to a human.
• Treat published accuracy figures as measured on a defined benchmark set, not as a guarantee for your documents. Only a threshold recorded in your agreement is contractual.

These are expectations about how the software should be used, not a warranty about outcomes. Where applicable law imposes stricter obligations on your use of automated processing, those obligations apply in addition.

8.Reporting a violation or a vulnerability

If you believe this policy is being breached, or you have found a security vulnerability in the software, the bundle, the documentation portal or the website, tell us through the contact page at https://bizfylabs.com/contact-us. Mark the message as a security report and we will route it to the product security team.

• Please include enough detail to reproduce the issue: the affected release and component, the environment, the steps taken and the observed result.
• Please give us a reasonable period to investigate and issue a fix before disclosing publicly, and we will keep you informed while we work.
• Do not include real personal data, customer documents or production credentials in a report. Use a synthetic or redacted example.
• We will not pursue a good-faith researcher who reports a vulnerability responsibly, stays within their own environment and does not access other parties' data.
• Where a vulnerability affects deployed releases, we publish an advisory and a fix to affected customers through the agreed technical contacts.

Existing customers may also raise a report through their normal support channel or their named commercial contact, whichever reaches us fastest.

9.Consequences of breach

How we respond depends on what happened. Most issues we have seen in evaluations are configuration mistakes or misunderstandings of a licence limit, and they are resolved by a conversation and a corrected order form.

• For a minor or inadvertent breach we will tell you what we believe has happened and ask you to remedy it within a reasonable period.
• Where a breach involves exceeding a deployment limit, we will agree a written variation to the order form covering the additional footprint.
• We may suspend support, a licence key or access to the documentation portal and private registry where continued use presents a serious legal, security or sanctions risk.
• We may terminate the licence for a material breach that is not remedied within the cure period in the terms and conditions.
• We may terminate immediately, without a cure period, for extraction or redistribution of model weights, for reverse engineering, or for use of the software for unlawful surveillance or the falsification of records.
• We reserve all other rights and remedies, including injunctive relief to protect our intellectual property and that of our licensors, and the indemnity in the terms and conditions.

Termination has the consequences set out in the terms and conditions: the software and the deployment bundle must be removed and all copies destroyed. Your documents and outputs stay with you, because they were never anywhere else.

10.Changes to this policy

We update this policy as the product changes, as the components we distribute change their own terms, and as the law on automated processing develops. The date at the top of the page shows the current version.

Where an update would materially restrict the use permitted under a licence already in force, it takes effect on renewal or through a written variation under the notice provisions of your agreement, not by a silent website change. Updates that clarify an existing restriction, or that reflect a change in applicable law, take effect when published.