SiRLab Synthetic Dataset License Agreement (v1.0)

This SiRLab Synthetic Dataset License Agreement (“Agreement”) is between SiRLab (“SiRLab”) and the person or entity accepting this Agreement (“Licensee”). “SiRLab” means the supplier identified as “SiRLab” on the applicable order form, invoice, or download page. By downloading, accessing, or using the Dataset (defined below), Licensee agrees to be bound by this Agreement.

1. Definitions

1.1 Dataset means the SiRLab-provided synthetic datasets (e.g., images, video, 3D scenes, point clouds, annotations, metadata, scripts, and documentation) identified at purchase or download.

1.2 Output means models, model weights, metrics, reports, visualizations, derived annotations, and other results generated by Licensee through processing or training using the Dataset, excluding the Dataset itself (in original or substantially similar form).

1.3 Authorized Users means Licensee’s employees and individual contractors (including service providers) working on Licensee’s behalf under confidentiality obligations at least as protective as this Agreement, and using the Dataset solely to support Licensee’s internal use for the Purpose.

For academic or research institutions, Authorized Users also includes enrolled students, visiting researchers, and external academic collaborators participating in a joint research project with Licensee, provided they (i) are bound by written obligations at least as protective as this Agreement and (ii) do not further redistribute the Dataset.

1.4 Confidential Information means non-public information disclosed by SiRLab that is marked or should reasonably be understood as confidential, including non-public Dataset components, tooling, and QA reports.

1.5 Affiliate means any entity that directly or indirectly controls, is controlled by, or is under common control with a party (“control” meaning ownership of more than 50% of the voting interests or the ability to direct management).

1.6 Item means one discrete sample unit as described in the Dataset card or documentation (e.g., one timestamp/frame/sweep for raw sensor data; and for annotations/labels, the corresponding annotation package for one such raw sample). Items are counted by logical entries, not by file boundaries; bundling multiple Items into a single file (or splitting one Item across multiple files) does not change the number of Items.

1.7 Manifest means the machine-readable inventory provided with the Dataset (or referenced on the download page) that identifies the Dataset contents, including item counts, data organization, and the Preview Files.

1.8 Dataset Version means the version identifier specified in the applicable Manifest (or, if not specified, the date/version shown on the Dataset card or download page for that Dataset).

1.9 Preview Files means any reduced, sample, or “preview” subset of the Dataset that SiRLab designates as preview/demo material (including sample frames, sample point clouds, or sample annotations) for display, documentation, or inspection purposes.

1.10 Secure Systems means information systems that implement reasonable administrative, technical, and physical safeguards appropriate to the sensitivity of the Dataset, including at minimum: (i) role-based access controls and least-privilege, (ii) unique user authentication with MFA for administrative access, (iii) encryption in transit using industry-standard TLS, (iv) encryption at rest (or compensating controls that provide equivalent protection), and (v) logging sufficient to support investigation of unauthorized access or disclosure.

1.11 Release Type means the designation in the applicable Manifest indicating whether a Dataset is provided as (a) Commercial (fee-based) or (b) Free (no-fee).

1.12 Commercial Dataset means a Dataset designated as “Commercial” in the Manifest (typically identified on an order form or invoice and subject to fees).

1.13 Free Dataset means a Dataset designated as “Free” in the Manifest, made available at no charge (e.g., public download for evaluation, research, or community use).

2. License Grant; Purpose

2.1 License. Subject to this Agreement and, for any Commercial Dataset, timely payment of any applicable fees, SiRLab grants Licensee a non-exclusive, non-sublicensable license that is non-transferable except as permitted under §18.1 and revocable only upon termination under §9 to download, install, reproduce, and use the Dataset solely for Licensee’s internal research, development, training, testing, and evaluation of AI/ML systems (including generative AI), and for creating derived synthetic data and annotations (the “Purpose”).

2.2 Model/Output Rights. Licensee may create, use, and commercialize Output (including trained models and services) provided that: (a) Output does not include the Dataset or any portion of it in a form that could serve as a substitute for the Dataset, and (b) Licensee does not redistribute the Dataset (see §4.4). For clarity, Output may include synthetic data, derived annotations, and datasets generated by Licensee (including via generative AI workflows), provided Licensee does not redistribute the Dataset itself except as permitted for Free Datasets under §2.6.

2.3 Affiliates / Contractors. Authorized Users may access the Dataset from Secure Systems (including Licensee-controlled systems or systems operated on Licensee’s behalf) for the Purpose. Licensee is responsible for their compliance and any breach by them is deemed a breach by Licensee. Affiliates. Licensee Affiliates may use the Dataset only if (i) they are identified on the applicable order form/invoice or (ii) they obtain a separate license from SiRLab. Service Provider Storage (Optional). Where operationally necessary for the Purpose, Licensee may permit individual contractors or service providers that qualify as Authorized Users to store and process copies of the Dataset on their own Secure Systems, provided that: (i) such party is bound by written obligations at least as protective as this Agreement, including non-redistribution and deletion/return upon completion; (ii) access is limited to personnel with a need to know; and (iii) Licensee maintains records of such storage locations under §7.2.

For academic or research collaborations, external academic collaborators who qualify as Authorized Users under §1.3 may also store and access copies of the Dataset on their Secure Systems solely for the joint research project, provided such access is subject to safeguards and confidentiality obligations at least as protective as this Agreement, and the Dataset is not further redistributed. For clarity, sharing the Dataset with such Authorized Users solely for the joint research project is not “redistribution” under §4.4.

2.4 Limited Excerpts. Licensee may reproduce and publicly display limited excerpts of the Dataset solely for documentation, academic publication, or presentations, provided such excerpts (i) are not a substitute for the Dataset, (ii) do not enable reconstruction of the Dataset, and (iii) include attribution per §15.

“Limited excerpts” means, per Dataset Version, no more than 5% of the total Items for each data type, rounded down (but in any event at least 1 Item per data type), subject to a maximum of 100 Items per data type. A minimum of 4 Items per data type applies only where that data type contains at least 80 total Items.

Counts are based on Items (as defined in §1.6) and the applicable Manifest (which specifies the applicable data types and item counts), not on the number of files; e.g., if multiple frames/sweeps are merged into one file, each frame/sweep is still considered a separate Item for purposes of this limit.

Data types are counted separately and include (by way of example) raw sensor data (e.g., camera images, LiDAR point clouds) and annotations/labels. For the avoidance of doubt, annotations/labels are treated as a separate data type category from raw sensor data, but all annotation modalities for a given sensor are counted together as one annotation data type (e.g., “camera annotations” includes 2D/3D bounding boxes, segmentation masks, tracks, attributes, calibration/poses, and metadata corresponding to camera data; “LiDAR annotations” includes analogous label/metadata associated with LiDAR data), even when such annotations correspond to the same underlying frame.

For clarity, if a raw sample Item is displayed publicly, any corresponding annotation package displayed with it counts separately toward the applicable annotation data type cap.

Licensee must not publish limited excerpts in a manner that enables bulk access, automated scraping, or dataset reconstruction (e.g., searchable indexes over excerpts, or releasing a large number of temporally contiguous frames that approximate full sequences).

2.5 Preview Files. Each Dataset includes (or is accompanied by) Preview Files for previsualization and inspection. Licensee may reproduce, redistribute, and publicly display Preview Files solely for Dataset cards, documentation, presentations, marketing, and inspection/purchase-decision purposes. Any redistribution or public display of Preview Files must clearly label them as ‘Preview Files’ and must not present results derived from Preview Files as benchmark, evaluation, or performance claims. Preview Files may be reduced and non-representative. Preview Files are part of the Dataset but are intended to preview structure/content only and may produce misleading or invalid performance metrics. Licensee must not use Preview Files (alone or combined with other data) as a substitute for the Dataset, nor use Preview Files for training, testing, validation, evaluation, or benchmarking of models or systems, nor to support performance claims or benchmark comparisons. For avoidance of doubt, loading, parsing, rendering, and schema/format validation of Preview Files is permitted where it is not used to generate performance metrics or comparative claims. Preview Files are included in the Dataset but are licensed only for the limited preview/inspection uses expressly permitted in this §2.5 and are otherwise excluded from the Purpose.

2.6 Free Datasets (Zero-Fee)

If the Dataset is a Free Dataset, then no fee is required to exercise the license granted under this Agreement. In addition, and notwithstanding §4.4, only where SiRLab expressly permits redistribution on the download page or in the Manifest, Licensee may redistribute the Free Dataset in unmodified form, provided that: (i) Licensee redistributes it at no charge (except reasonable cost of media or bandwidth), (ii) Licensee includes with any redistribution (a) a copy of this Agreement and (b) the applicable Manifest, and provides a prominent attribution notice in documentation and any download page in substantially the following form:

SiRLab Synthetic Dataset — Dataset Version [X] (Manifest: [ID/date]). © SiRLab. Used under SiRLab Synthetic Dataset License Agreement. Source: https://sirlab.ai/legal/dataset-license/latest

(iii) all proprietary notices, attributions, and identifiers remain intact, and (iv) the redistribution does not imply endorsement by SiRLab.

Licensee may create, use, and commercialize Output created using a Free Dataset under the same conditions as §2.2.

For clarity, a Free Dataset remains subject to all applicable restrictions in §4 (including prohibited use cases), and any additional terms stated in the Manifest for that Free Dataset.

Licensee may redistribute modified versions of a Free Dataset only if (a) the modified dataset is clearly marked as modified, (b) it is not branded as an official SiRLab dataset, and must be distributed under a distinct name that does not include ‘SiRLab’ (or any confusingly similar mark), (c) it includes prominent attribution to SiRLab as the original source, and (d) it does not include any portion of the Dataset in a way that violates §4 or enables Dataset reconstruction beyond what is already present in the Free Dataset.

3. Confidentiality

3.1 Licensee will protect Confidential Information using at least reasonable care and use it only for the Purpose. Disclosure is limited to Authorized Users with a need to know and bound by written confidentiality obligations.

3.2 “Confidential Information” includes non-public Dataset details (e.g., QA methods, labeling schemes not published in the dataset card) and any non-public evaluation materials provided by SiRLab, except where such information is independently and lawfully obtained or publicly available without breach. Licensee’s independently produced benchmarks and results are not Confidential Information unless they include or disclose SiRLab’s non-public materials.

3.3 Academic Publication. Notwithstanding the above, Licensee may disclose Dataset characteristics and label taxonomies to the extent reasonably necessary for academic or scientific publication, provided Licensee does not disclose SiRLab’s non-public QA reports, tooling, or unreleased Dataset components, and remains subject to the Limited Excerpts rule in §2.4.

4. Restrictions

Licensee must not:

1. Use the Dataset for unethical surveillance or to directly enable law enforcement or public authorities to monitor individuals; or to identify, track, or profile individuals, vehicles, or groups in real-world footage.
2. Attempt to infer or process sensitive attributes (e.g., race, ethnicity, health) or perform biometric identification, emotion recognition, or social scoring.
3. De-anonymize, re-identify, or link the Dataset to any natural person or real-world plate/ID; attempt to match Dataset elements to real persons, vehicles, or properties.
4. Redistribute, resell, sublicense, publish, host, embed, or make the Dataset available (in whole or part) to any third party, including as part of a product, service, dataset, benchmark, model card repository, or API, or in any manner that allows a third party to download, extract, query, or retrieve the Dataset (or any portion of it). (Permitted: sharing Output as per §2.2 and sharing with Authorized Users as per §2.3.) For clarity, this restriction does not prohibit hosting or providing Output (including models/services) as a service, provided the Dataset itself is not made available or exposed. Exception: This restriction does not apply to Preview Files as permitted under §2.5, or to Free Datasets to the extent redistribution is expressly permitted under §2.6.
5. Remove proprietary notices or bypass technical controls.
6. Use the Dataset outside the Purpose without SiRLab’s prior written consent. This clause does not permit any use that is prohibited under this Agreement or applicable law.
7. Use the Dataset to train or evaluate identity recognition systems, emotion inference systems, or social credit scoring systems.
8. Use Preview Files for training, testing, validation, evaluation, or benchmarking, or to support performance claims or benchmark comparisons.

5. Data Protection; Ethics

5.1 Synthetic data statement. SiRLab designs the Dataset to be synthetic and not to contain personal data. If Licensee believes any element may constitute personal data, Licensee must promptly notify SiRLab at contact@sirlab.ai and cease such use until SiRLab advises.

5.2 GDPR/Compliance. Licensee remains responsible for compliance with applicable laws, including EU GDPR where relevant to Licensee’s use or combination of the Dataset with other data.

5.3 Responsible AI. Licensee should mitigate bias, document limitations, and disclose appropriate usage notes to downstream users of Output.

5.4 Security. Licensee will implement reasonable administrative, technical, and physical safeguards consistent with Secure Systems, designed to prevent unauthorized access to or redistribution of the Dataset, including access controls and least-privilege principles. Licensee will notify SiRLab without undue delay after becoming aware of any unauthorized access to or disclosure of the Dataset and will cooperate reasonably to remediate and prevent recurrence.

6. Ownership; Feedback

6.1 SiRLab and its licensors retain all rights, title, and interest in and to the Dataset and associated IP. No rights are granted by implication.

6.2 Licensee owns its Output, subject to SiRLab’s rights in the Dataset.

6.3 Feedback. If Licensee provides feedback, SiRLab may use it without restriction and without any obligation to Licensee.

7. Delivery; Updates; Records

7.1 Delivery is by electronic download/access. SiRLab may provide updates or replacements and may request that Licensee delete or retire prior versions when security, legal, or quality requires it. Where reasonably practicable, SiRLab will provide Licensee with notice and a replacement version or mitigation guidance.

7.2 Licensee will maintain reasonable internal records of locations where copies of the Dataset are stored to support compliance and deletion requests.

7.3 Manifest; Previews. Each Dataset is identified by its applicable Manifest, which specifies the Dataset Version and may include or reference Preview Files. Where this Agreement refers to item counts or Dataset Version, the Manifest controls.

8. Fees; Taxes

Fees (if any) are as agreed at order. Some Datasets may be designated as Free Datasets in the Manifest and are provided at no charge. Fees are non-refundable unless expressly stated. Prices are exclusive of taxes; Licensee is responsible for applicable taxes, excluding SiRLab’s income taxes.

9. Term; Termination

9.1 Term. This Agreement starts upon download/access and continues unless terminated as below.

9.2 Termination for Cause. SiRLab may terminate immediately if Licensee breaches §4.4 (redistribution/making available), §13 (export/sanctions), or commits a willful or repeated breach of §§3–5 or §4. For other material breaches, SiRLab may terminate if Licensee fails to cure within 10 days of notice (or 30 days where the breach is not reasonably curable within 10 days, provided Licensee is diligently pursuing cure). Licensee may terminate at any time by ceasing use and deleting the Dataset.

9.3 Effect. Upon termination, Licensee must stop all use and delete all copies of the Dataset and certify deletion upon request. Outputs lawfully created before termination may be retained and used (subject to §2.2), provided Licensee remains in compliance with surviving terms.

9.4 Survival. §§3–6, 8–18 survive termination.

10. Warranties; Disclaimer

The Dataset is provided “AS IS”. To the maximum extent permitted by law, SiRLab disclaims all warranties, whether express, implied, or statutory, including non-infringement, merchantability, fitness for a particular purpose, and accuracy or completeness of labels or metadata. The Dataset may contain errors or artifacts and is intended for engineering use by skilled professionals.

Right to License. SiRLab represents that it has the right to grant the license in §2 to the Dataset as provided. This is Licensee’s sole and exclusive remedy for any breach of this sentence.

11. Limitation of Liability

To the maximum extent permitted by law, SiRLab will not be liable for any indirect, incidental, special, consequential, exemplary, or punitive damages, or for lost profits, revenues, goodwill, or data, arising from or related to this Agreement or the Dataset, even if advised of the possibility. SiRLab’s aggregate liability under this Agreement will not exceed the amounts paid by Licensee for the Dataset giving rise to the claim in the 12 months preceding the claim.

12. Indemnity

To the extent permitted by applicable law, Licensee will defend, indemnify, and hold harmless SiRLab and its personnel from and against third-party claims, damages, liabilities, costs, and expenses (including reasonable attorneys’ fees) arising out of Licensee’s (a) use of the Dataset in violation of this Agreement or law, (b) redistribution or making available of the Dataset, or (c) Outputs that infringe rights or cause harm.

For clarity regarding clause (c), Licensee’s indemnity applies to the extent the claim arises from Licensee’s combination of the Dataset with other data, modifications, training choices, deployment, or use outside the Purpose or in violation of law.

This §12 does not apply to the extent a claim arises solely from the unmodified Dataset as provided by SiRLab, except where Licensee has combined the Dataset with other data, modified it, used it outside the Purpose, or violated this Agreement.

If Licensee is a public academic or governmental institution, the obligations in this §12 apply only to the extent permitted by applicable law; to the extent indemnification is prohibited, each party remains responsible for its own acts and omissions.

13. Export; Sanctions; Safety

Licensee will comply with applicable export control and sanctions laws (including EU and U.S. laws). Licensee will not use the Dataset for prohibited end uses (e.g., WMD development) or provide it to restricted parties or destinations. If required by law or safety concerns, SiRLab may suspend access.

14. Audit (Light-Touch)

Upon reasonable written notice (and no more than once per 12 months), Licensee will provide a written certification of compliance signed by an officer. SiRLab may request reasonable additional information limited to verifying (i) storage locations, (ii) access controls, and (iii) non-redistribution. Any information provided under this §14 will be treated as confidential information of Licensee. SiRLab will protect such confidential information of Licensee using at least reasonable care and will use it only to verify compliance under this §14. SiRLab will not request access to Licensee source code or production systems under this §14.

15. Publicity; Attribution

SiRLab may list Licensee’s name and logo as a Dataset customer only with Licensee’s prior written consent. If Licensee publishes academic work using the Dataset, Licensee will include an appropriate citation to “SiRLab Synthetic Dataset” and the Dataset card/version. Except for the attribution required by this Agreement, no license is granted to SiRLab’s trademarks or logos, and any use must not imply endorsement.

Licensee may publish academic or scientific results (including benchmark tables) that reference the Dataset, provided such statements are factual and include the Dataset card/version. Use of SiRLab’s name or logo for marketing, endorsement, or comparative advertising requires SiRLab’s prior written consent.

16. Governing Law; Venue; Language

This Agreement is governed by the laws of Belgium, without regard to conflict-of-laws rules, and subject to any mandatory provisions of applicable European Union law. For B2B relationships, the courts of Brussels, Belgium have exclusive jurisdiction, and the parties consent to personal jurisdiction there. For consumers (B2C) habitually resident in the European Union, nothing in this Agreement limits any mandatory consumer rights regarding jurisdiction or applicable law under EU rules. Any final judgment of the Brussels courts may be recognised and enforced across EU Member States in accordance with applicable EU rules on jurisdiction and the recognition and enforcement of judgments in civil and commercial matters.

Language. This Agreement may be made available in French. If a French version is provided at acceptance, it controls. In the event of any inconsistency, the parties will seek a common interpretation in good faith consistent with Belgian law; failing agreement, the courts identified above will determine the interpretation.

17. Notices

Legal notices to SiRLab: legal@sirlab.ai. Notices to Licensee may be sent to the email/address provided at purchase.

18. Miscellaneous

18.1 Assignment. Licensee may not assign this Agreement without SiRLab’s prior written consent; any prohibited assignment is void. Notwithstanding the foregoing, Licensee may assign this Agreement without consent to an Affiliate or in connection with a merger, acquisition, corporate reorganization, or sale of substantially all of its assets, provided the assignee agrees in writing to be bound by this Agreement. SiRLab may assign in connection with a reorganization or sale.

18.2 Force Majeure. Neither party is liable for delays caused by events beyond reasonable control.

18.3 Severability; Waiver. Unenforceable terms will be modified to the minimum extent needed; failure to enforce is not a waiver.

18.4 Entire Agreement; Order of Precedence. This is the entire agreement concerning the Dataset and supersedes prior communications. If an order form conflicts with this Agreement, the order form controls for commercial terms (fees, quantity, SKU), and this Agreement controls for legal terms.

18.5 Amendments. Amendments must be in writing and signed by both parties.

By downloading or using the Dataset, Licensee agrees to this Agreement.