Privacy Policy

The sections below explain collection, use, sharing, retention and your rights.

1. Who We Are

Om Therapeutics, Inc., a Delaware C Corporation with its principal place of business at 3210 Merryfield Row, San Diego, CA 92121 USA ("Om," "we," "us") provides software, APIs, and related services that incorporate machine-learning ("ML") techniques. This Privacy Policy ("Policy") explains our global data-handling practices.

2. Scope

3. Key Definitions

4. Information We Collect

5. How and Why We Use Data

Lawful bases for EU/UK data subjects are detailed in Section 6.

5.1 Scientific data and model improvement

We process Customer Inputs and Customer Results to perform, deliver, support, secure, troubleshoot, and validate your requested Services, including private analysis or fine-tuning you expressly request. Customer scientific data and learned updates are private by default. Om may use them or learning derived from them to train or improve its own or shared models, methods or services only when you explicitly approve the identified data and uses through an opt-in path. Such paths may be marketed as Open Pipeline, Discovery Marketplace or other program names. Accepting general Terms, using Services or participating in a program does not by itself authorize training; renaming a program does not expand approval. Each approval identifies the covered data and version or defined forthcoming result, permitted uses, applicable terms, duration and survival or withdrawal conditions. Complete screens, assay controls, positive, negative and inconclusive results, provider work, source datasets and private fine-tunes are not automatically included. Approval covers only its identified scope, not unrelated results or future work. Derived labels, features, learned parameters, evaluation, benchmarking, checkpoint averaging, merging and distillation retain the same restrictions. Requested private processing does not authorize shared improvement. Model-use survival follows your explicit approval of the uses, model/artifact scope and duration; approval to train alone does not grant indefinite model use. Continued model use does not authorize new training or retention/reuse of source records unless separately approved. Partnered Pipeline uses and closeout follow Section 5.6 of the Terms and all contributing-source restrictions. Om protects non-public Customer Inputs and Customer Results and limits access to authorized personnel and providers bound by confidentiality and applicable use restrictions. Permitted improvement does not authorize Om to publish, sell, license, or disclose your customer-specific raw results, original training dataset, or confidential project details to other customers as a dataset or data product. Such disclosure requires your express written permission or specific program terms you explicitly approve for that disclosure. We use reasonable safeguards against exposing confidential source records through model outputs. The improvement permission does not authorize using your private inputs, raw or processed results, candidate lists, or customer-specific fine-tuned model to carry out separate research for Om or another customer. Om may use resulting eligible shared models in otherwise permitted work, with safeguards against retrieving or repackaging private results. Separate marketplace enrollment may expressly authorize use of identified enrolled molecules and data in Om's Open Pipeline research and discovery under Section 6.9 of the Terms. Training approval alone does not grant that research permission, and unenrolled private material remains protected. Outside an expressly approved research permission, Section 5.3's non-exclusive same-target work must rely on independently available information Om has the right to use; Om must substantiate those sources. Your confidential submission is not a general license to select or pursue a target. Public knowledge of the target does not make your involvement, rationale, modifications, private inputs or results public. Separate NDAs and historical restrictions still apply. Work performed to fulfill your Services remains customer-owned. Work on the same or similar sequence, target, molecule, or workflow can produce similar results. It does not authorize another customer to access your raw results. The private default applies without an enterprise upgrade. Signed agreements may add restrictions on covered material and descendants while permitting requested private processing. Sequence exclusivity requires a separate express agreement on scope and duration; within that scope and duration Om will not perform or commission the covered competing activities for itself or third parties, subject only to express signed exceptions. Exclusivity does not itself grant training permission; any explicitly approved improvement must still respect the covered-activity restriction. For designated Partnered Pipeline programs, any surviving use requires explicit approval and the closeout duties below apply. Specific Open Pipeline Challenge acquisition, payout, training, and publication conditions control Challenge material; general Terms acceptance does not bypass them or grant training rights. Express marketplace enrollment under Section 6.9 of the Terms is a separate election. On the website or through a supported MCP client, the customer's single explicit confirmation of a reviewed listing constitutes that election and approval of the disclosed Marketplace, model-training and Open Pipeline research uses. No additional legal or training approval is required for the same reviewed listing. Confirmation of a purchase accepts its buyer terms but does not enroll the buyer's own data. Initial general Terms and Privacy agreement remains in the website flow. The contributor retains ownership and may use enrolled data privately or publish generalized findings that do not disclose or permit reconstruction of the exact enrolled structure or Marketplace Data, while granting Om exclusive marketplace hosting and buyer-licensing rights. Marketplace enrollment is the only owner sharing election available through the Services; new free-publication elections are unavailable. Historical records made public before retirement of that feature remain readable under their prior terms. As conditions of hosting, the contributor must explicitly approve both a perpetual, non-exclusive Open Pipeline research permission and a perpetual, non-exclusive license to use the specifically enrolled data and molecular information for model training, retraining, fine-tuning, evaluation and improvement. The research permission allows Om and its providers to use the enrolled molecules and data to select and pursue Open Pipeline programs and procure, synthesize, screen, test, evaluate and develop those molecules for research, within the rights the contributor can authorize and subject to signed restrictions and third-party rights. It does not grant a patent license or public-release rights. Both permissions must be disclosed prominently in the enrollment review and begin upon accepted enrollment, independently of a sale or Challenge payout. Browsing or buying does not enroll a customer's work. Earlier enrollments retain their accepted scope; new research rights require fresh explicit approval. Neither permission activates the conditional Challenge license, changes an award obligation or covers unenrolled data. Prior rights and signed restrictions remain controlling. Personal information is excluded even when embedded in an enrolled scientific record.

5.2 Personal information and support

We do not use personal information, account information, credentials, or payment data for model training or sale. This exclusion also applies when personal information appears in scientific files or support communications. We may use support communications to provide support, troubleshoot, and improve support quality and product features. We do not sell support communications or treat them as scientific training data merely because you send them. We use reasonable data minimization, access controls, and appropriate retention practices. Any personal-information processing remains subject to applicable law.

6. Legal Bases (EU/EEA/UK GDPR and Comparable Laws)

Contractual permission to use scientific data is separate from the lawful basis for processing personal information. Including improvement in a contract does not by itself make personal-information processing necessary to perform that contract. We rely on contractual necessity only where processing is objectively necessary to deliver the requested service. Our legitimate interests include maintaining useful, reliable products and support, subject to the required assessment and your rights. We obtain consent where required. Personal information remains excluded from model training under Section 5.2.

7. Disclosures to Third Parties

8. Your Rights and Choices

8.1 Account closure and scientific data

You retain ownership of your inputs, customer-specific results, and downstream work after account closure or subscription cancellation. There is no default shared-improvement license. Any expressly approved training permission continues or ends as stated in its applicable opt-in terms, subject to law and signed restrictions. Closure itself grants no rights. Ownership, future training permission, operation of already trained models, service storage, and personal-information deletion are separate matters. The distinct hosting, buyer-licensing, model-training and Open Pipeline research permissions expressly granted by marketplace enrollment survive listing suspension and account closure as stated in the Terms. Their permanence concerns the enrolled scientific material, not a waiver of personal-information rights. Open Pipeline's perpetual scientific-data licenses and permanent enrollment do not remove your statutory rights in personal information. We retain necessary purchase, license, historical-publication, access-grant, contributor-earnings and payout records to administer continuing rights and unpaid amounts, handle refunds or disputes, prevent fraud and meet applicable legal or accounting obligations. We limit personal information to what those purposes require and apply the retention principles in Section 11. A deletion request does not automatically erase required transaction records or another buyer's scientific license; we assess the request under applicable law. Account closure does not make personal, account, identity-verification or payment data available for model training or sale. This Policy does not retroactively expand rights in previously provided or generated material. Earlier agreements and privacy commitments, including any promise to stop future training upon account closure, continue to govern that material unless expressly amended. Shared-improvement exclusions continue for covered work and its descendants as agreed; account closure does not remove them. Sequence exclusivity follows its separately agreed duration and payment conditions. Expiry of exclusivity does not remove confidentiality or a surviving data-use exclusion. Deletion does not itself promise removal from models already lawfully trained, and any legal or contractual obligation requiring that outcome still applies. For designated Partnered Pipeline programs, continued improvement or model use after the program ends follows only the scope and duration you expressly approve, including withdrawal and termination conditions. Without that approval, shared improvement ends with the program and affected models must be retired from use for Om and other customers across all targets. At program end or earlier expiry or withdrawal of a permission, Om stops every use no longer authorized and applies agreed closeout to covered inputs, results, derived material and all affected models, including shared and merged descendants. Approved surviving uses remain subject to confidentiality, exclusivity while in force, and every contributing source's restrictions. They do not authorize source-record retention or new training unless separately approved, or silently override prior signed retirement or deletion duties. Orders placed and accepted during the program will still be honored. Om and its providers may process covered material and affected models only as needed to complete and deliver those orders for that customer, including scientific work expressly ordered during the term. Results and learned artifacts generated during completion remain covered partner material; any shared improvement requires applicable explicit approval covering those results. Completion itself does not extend exclusivity or an improvement permission. Om may retain covered material and affected artifacts in a restricted, inactive archive to preserve the program for a possible restart of that customer's partnership. Apart from accepted-order completion or a specifically approved surviving use, active scientific use must stop. Archived copies cannot be used for research, training, evaluation, inference, model combination or other scientific or commercial activity. Only handling needed to secure and maintain the archive, provide agreed export or return, or comply with legal duties and required deletion is permitted. The same restrictions apply to providers, copies and backups. Partner fine-tuned models and their weights, checkpoints and other learned-parameter files remain on Om-managed infrastructure under Section 7.3 of the Terms, including during closeout. Agreed export or return under the partnered program covers Customer Inputs and non-model Customer Results, not those model artifacts. Separately licensed delivery and downloaded-copy treatment require the written commercial model-license agreement described in Section 7.3 of the Terms. The signed closeout schedule specifies existing-order handling, any approved surviving uses and their model/data scope, duration and cessation duties separately from source-record retention, export and archival deadlines, written confirmation, retention periods or review criteria, and restart conditions. Personal-information retention must remain necessary and lawful; applicable deletion rights and stricter historical or signed obligations continue. Archiving does not permit indefinite retention of personal information merely because a partnership might restart. Reactivation requires a signed restart or renewal of the same customer's Partnered Pipeline program identifying the archived material and expressly authorizing renewed use, subject to all applicable source restrictions. Remaining credits, account reopening or later self-service orders do not reactivate it. Om must track data and model dependencies and control copies to enforce these limits. Outside a specifically approved surviving use, accepted-order completion or signed restart, later Om work requires data and models that never incorporated the covered material or its learning. Continued model-use approval does not reactivate an inactive archive. Customer ownership and confidentiality continue. Section 5.6 of the Terms controls.

8.2 Enterprise choices and service access

Private-by-default treatment requires no paid upgrade or signed exclusion. Additional data-use restrictions and sequence exclusivity may be negotiated through a signed enterprise agreement. Contact support@omtx.ai to discuss scope before submitting covered work. Standard confidentiality and ownership apply to all customers. Your statutory personal-information rights are independent of purchasing enterprise terms. A signed agreement may include exclusivity in a recurring enterprise spending commitment; any scientific-data improvement permission still requires your explicit approval of the identified data and uses. A designated Partnered Pipeline program follows the explicitly approved duration and survival conditions and requires the remaining data and model closeout described above. These protections take effect only as stated in the signed agreement. Where the agreement ties exclusivity to a cancel-anytime monthly commitment, exclusivity ends at the close of the paid service period after cancellation unless expressly agreed otherwise. Unused credits do not extend it. Confidentiality and customer ownership continue. Subscriptions, Wallet Credits, delivery and export access are governed by the applicable service terms; ownership does not provide perpetual hosting. All standard hosted fine-tunes, including self-service and partnered fine-tunes, are stored and run on Om-managed infrastructure, including cloud infrastructure operated for Om. Access is through Om-authorized hosted Services. Standard fine-tuning does not include delivery, download, export or transfer to you or your designees of weights, checkpoints, adapters, deltas or other learned-parameter files, including customer-owned updates. Any customer-specific model delivery or off-platform use requires a separately negotiated written commercial model-license agreement signed by you and Om and separately agreed delivery arrangements. Standard fine-tuning includes no model-download facility or commitment to provide one. The separate agreement governs identified artifacts, permitted uses, downloaded copies, duration and closeout, subject to unamended signed duties and other contributing sources' restrictions. This does not change ownership, permit shared training, or limit agreed scientific-result exports. Storage and retention remain subject to applicable schedules and deletion duties. Enterprise partnerships include Om's commitment to use commercially reasonable efforts to onboard requested new capabilities. Work through those capabilities and sharing with Om-engaged providers remain subject to the same applicable ownership, confidentiality, and data-use protections.

9. Security

We use reasonable administrative, technical, and organizational safeguards designed to protect personal information. No method of transmission or storage is completely secure.

10. International Transfers

We may transfer personal information across borders. Where required, we rely on appropriate safeguards such as standard contractual clauses or other lawful transfer mechanisms.

11. Data Retention

We retain personal information for as long as needed to provide the Services, comply with legal obligations, resolve disputes, and enforce agreements. Retention periods may vary by data type and jurisdiction. Scientific deliverables, export availability, and any storage commitment are specified by the product or order. Om may retain permitted scientific material for improvement only within the scope and duration you explicitly approved, or as permitted by its governing historical agreement; this does not extend personal-information retention automatically. Required deletion, restricted archival copies, and any earlier account-closure obligations remain applicable. For Partnered Pipeline, the specific closeout duties in Section 8.1 override general retention for continuing improvement or commercial purposes. Restricted scientific archives may preserve the program for a possible agreed restart, subject to the signed retention periods or review criteria and applicable law; their existence does not authorize active use or override personal-information deletion rights.

12. Children's Privacy

The Services are not directed to children under 18, and we do not knowingly collect personal information from children.

13. Changes to This Policy

We may update this Policy from time to time. We will email a summary and links to material updates before their stated effective date as described in Section 15 of the Terms, and may also display a notice in the Services. Routine policy notices do not require repeated acceptance or interrupt existing workflows. This Policy explains our practices; notice, continued use, or acceptance of the Terms does not substitute for any specific consent required by applicable privacy law. Where required, we obtain that consent before the affected processing. Updates do not silently change existing signed agreements, specific program terms, historical data rights, or the signed requirements for reactivating archives.

14. Contact