AI Terms Exhibit

Version 1.0 — Effective 2026-09-09 — incorporated into the SummaGuard Subscription Terms v1.0

SummaGuard is a product of SummaCore LLC

This AI Terms Exhibit (this "Exhibit") is incorporated into and forms part of the SummaGuard Subscription Terms (the "Agreement") between SummaCore LLC, a Texas limited liability company ("SummaCore," "we," "us"), and the customer identified in the applicable Order ("Customer," "you"). This Exhibit governs Customer's access to and use of the AI Features (defined below) included in the Service. Capitalized terms not defined in this Exhibit have the meanings given in the SummaGuard Subscription Terms. If this Exhibit conflicts with the SummaGuard Subscription Terms with respect to the AI Features, this Exhibit controls.

1. Scope and Definitions

1.1 Scope. This Exhibit applies to Customer's use of the AI Features made available as part of the Service at guard.summacore.com, including the AI assistant marketed as "eva."

1.2 Definitions.

1.3 Inputs and AI Outputs are Customer Data. Inputs and AI Outputs constitute Customer Data under the Agreement and are protected by the confidentiality, security, and data-processing commitments of the SummaGuard Subscription Terms, the Data Processing Addendum, and the Security Exhibit.

2. How the AI Features Work

2.1 Retrieval-grounded operation. The AI Features are designed to be retrieval-grounded: when eva answers a question or drafts content, it retrieves relevant passages from Customer's indexed document library within the Service, and loads the specific records the User is working in. AI Outputs include citations to the documents retrieved, where applicable, so Customer can verify the source of any statement.

2.2 Suggest-then-confirm. The AI Features operate on a suggest-then-confirm basis. AI Outputs are presented as drafts or suggestions for human review. The AI Features do not autonomously create, modify, finalize, or submit records, filings, or determinations; a human user must review and confirm any action before it takes effect in the Service, except that the AI Features may create clearly labelled drafts and derived translations; a draft or translation has no effect on any record, filing, or determination until a User reviews and confirms it, and the AI Features cannot activate or finalize anything. Actions confirmed by a user (or taken by an automated client under Customer's Access Credentials, as described in the SummaGuard Subscription Terms) are recorded in the Service's audit trail.

2.3 SummaCore-controlled infrastructure. All AI inference for the AI Features runs on computing infrastructure owned or controlled by SummaCore. The AI Features do not send Customer Data to any third-party cloud AI or large-language-model provider, and SummaCore uses no third-party AI sub-processors in the delivery of the AI Features. The Data Processing Addendum lists all sub-processors used to provide the Service; none of them receives Customer Data for AI processing.

2.4 Nature of generative AI. The AI Features use machine-learning models that generate probabilistic output. AI Outputs may be inaccurate, incomplete, out of date, or inappropriate for Customer's specific circumstances, even when citations are provided. Similar Inputs may produce different AI Outputs at different times. This is inherent to the technology and is why the verification obligations in Section 4 apply.

3. No Training on Customer Data

3.1 Unconditional commitment. SummaCore does not use Customer Data — including Inputs, AI Outputs, usage of the AI Features, or any data derived from Customer Data, whether identifiable, de-identified, aggregated, or anonymized — to train, retrain, fine-tune, tune, calibrate, or otherwise improve any artificial-intelligence or machine-learning model, whether SummaCore's or anyone else's. This commitment is unconditional and has no carve-outs.

3.2 No third-party exposure. Because AI inference runs entirely on SummaCore-controlled infrastructure with no third-party AI sub-processors (Section 2.3), Customer Data is not made available to any third-party model provider for any purpose, including training, by construction of the Service.

3.3 Relationship to service operation. Nothing in this Section 3 limits SummaCore's ability to process Customer Data as needed to provide the AI Features to Customer at Customer's direction (for example, retrieving and processing Customer's documents to answer Customer's question), consistent with the Agreement and the Data Processing Addendum. Providing the Service is not training.

4. Customer Responsibilities

4.1 Verification of AI Outputs. AI Outputs are drafts and suggestions only. They are not a substitute for Customer's professional judgment or for review by qualified personnel. Customer is solely responsible for reviewing and verifying the accuracy, completeness, and appropriateness of all AI Outputs before using or relying on them, and for all decisions made and actions taken in reliance on AI Outputs.

4.2 Human review for regulated determinations. Without limiting Section 4.1, Customer must ensure that a qualified human reviews and approves any AI Output before it is used in connection with: (a) OSHA recordability or work-relatedness determinations; (b) incident or near-miss investigations and root-cause analyses; (c) OSHA Forms 300, 300A, and 301, Injury Tracking Application submissions, and any other regulatory filing or submission; (d) privacy-case designations under 29 CFR 1904.29(b)(6)–(9); and (e) any determination affecting employee health, safety, discipline, or employment. The AI Features support and assist these activities; they do not perform or determine them.

4.3 Responsibility for Inputs. Customer is responsible for its Inputs, including ensuring that Customer has all rights and consents necessary to submit them to the Service and that they comply with applicable law and the Acceptable Use Policy. Customer should not submit Inputs seeking output the AI Features are not designed to provide (for example, medical diagnoses or legal opinions).

4.4 Compliance remains Customer's obligation. As stated in the SummaGuard Subscription Terms, Customer remains solely responsible for its own compliance with OSHA regulations and all other laws applicable to Customer. Use of the AI Features does not transfer, reduce, or satisfy any legal obligation of Customer.

5. Ownership

5.1 Customer owns AI Outputs. As between the parties, Customer owns the AI Outputs generated for Customer, and AI Outputs are Customer Data under the Agreement. To the extent SummaCore holds any right, title, or interest in AI Outputs, SummaCore assigns it to Customer upon generation.

5.2 No rights in models. Except for the express right to use the AI Features as part of the Service during the Subscription Term, no right, title, or interest in the AI Features, the underlying machine-learning models, model weights, software, or infrastructure is granted to Customer. Customer will not (and will not permit anyone to) reverse engineer, extract, or attempt to replicate the underlying models, or use the AI Features to develop a competing product or service.

5.3 Non-uniqueness. Due to the nature of machine learning, AI Outputs may be similar to output generated for other customers from similar prompts against different data. Customer's ownership under Section 5.1 does not extend to output independently generated for other customers.

6. Warranty Disclaimer and Limitation of Liability for AI Outputs

6.1 Disclaimer. TO THE MAXIMUM EXTENT PERMITTED BY LAW, THE AI FEATURES AND ALL AI OUTPUTS ARE PROVIDED "AS IS" AND "AS AVAILABLE," WITHOUT WARRANTY OF ANY KIND. SUMMACORE MAKES NO REPRESENTATION OR WARRANTY, EXPRESS, IMPLIED, STATUTORY, OR OTHERWISE, THAT AI OUTPUTS WILL BE ACCURATE, COMPLETE, CURRENT, RELIABLE, ERROR-FREE, OR FIT FOR ANY PARTICULAR PURPOSE, OR THAT USE OF THE AI FEATURES WILL SATISFY ANY LEGAL OR REGULATORY REQUIREMENT, INCLUDING ANY REQUIREMENT UNDER 29 CFR PART 1904 OR ANY OTHER OSHA REGULATION. ALL IMPLIED WARRANTIES, INCLUDING MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT, ARE DISCLAIMED WITH RESPECT TO THE AI FEATURES AND AI OUTPUTS.

6.2 Exclusion of liability for reliance. TO THE MAXIMUM EXTENT PERMITTED BY LAW, SUMMACORE WILL HAVE NO LIABILITY FOR ANY LOSS, DAMAGE, CLAIM, PENALTY, CITATION, OR FINE ARISING FROM OR RELATING TO CUSTOMER'S USE OF OR RELIANCE ON ANY AI OUTPUT, INCLUDING ANY DECISION, DETERMINATION, CLASSIFICATION, FILING, OR OMISSION MADE IN RELIANCE ON AN AI OUTPUT. THIS SECTION 6.2 OPERATES IN ADDITION TO, AND DOES NOT LIMIT, THE EXCLUSIONS AND LIMITATIONS OF LIABILITY IN THE SUMMAGUARD SUBSCRIPTION TERMS, INCLUDING THE AGGREGATE LIABILITY CAP.

eva and the other AI Features provide information and drafting assistance only. They do not provide legal, medical, industrial-hygiene, engineering, or other professional advice, and no AI Output constitutes such advice or creates any professional-client relationship. The AI Features support, assist with, and draft materials for Customer's review; they do not determine, ensure, or guarantee OSHA compliance, recordability, or compliance with any law. Recordability determinations, OSHA filings, and regulatory compliance remain Customer's responsibility (Section 4). Customer should consult its own qualified legal, safety, and medical professionals regarding its obligations.

8. Availability of AI Features

The AI Features are provided on an as-available basis and run on dedicated AI infrastructure separate from the core recordkeeping functions of the Service. The AI Features may be degraded, throttled, or unavailable from time to time without notice. AI Features are provided on an as-available basis and are not subject to any availability commitment. The core recordkeeping functions of the Service (including incident records, OSHA forms, and exports) are not dependent on the availability of the AI Features. SummaCore may modify, replace, or discontinue individual AI Features, provided the modification does not materially reduce the overall functionality of the Service during a paid Subscription Term.

9. Responsible AI

9.1 Informed by the NIST AI RMF. SummaCore's AI risk-management program is informed by the National Institute of Standards and Technology Artificial Intelligence Risk Management Framework (NIST AI RMF 1.0), covering the governance, mapping, measurement, and management of risks associated with the AI Features, as further described in the Security Exhibit.

9.2 Audit logging. AI Features that create or persist a record are recorded in the Service's audit trail with the requesting User and a timestamp. SummaCore does not record the content of prompts or AI Outputs in the audit trail.

9.3 Design controls. The AI Features are designed with the controls described in Section 2 — retrieval grounding over Customer's own data, source citations, suggest-then-confirm human review, and inference confined to SummaCore-controlled infrastructure — as standing product-level safeguards. Additional technical and organizational measures applicable to the AI Features are set out in the Security Exhibit.

End of AI Terms Exhibit — Version 1.0 — Effective 2026-09-09 — incorporated into the SummaGuard Subscription Terms v1.0