AI Terms

1. Relationship to Agreement. These AI Terms (“AI Terms”) are incorporated into the Agreement. Capitalized terms not defined in this Addendum shall have the meanings ascribed to them in the Agreement. Except as expressly provided herein, nothing in these AI Terms shall be deemed to waive or modify any of the provisions of the Agreement. In the event of a conflict between a provision of these AI Terms and the Agreement, these AI Terms will control. Otherwise, these AI Terms will be construed under, and in accordance with, the terms of the Agreement.

2. Artificial Intelligence Usage.

2.1 AI Definitions.

“AI” means artificial intelligence.

“AI Assets” means (i) Archipelago models, algorithms, model weights, training methodologies, and proprietary enrichment schema used in the Service, regardless of whether these were improved based on Customer Data or Customer usage; (ii) validation rules and learned plausibility patterns; enrichment logic, including document extraction rules, field mapping rules, and gap-fill inference logic and statistical benchmarks; and (iii) Derived Data, Benchmark Data, AI Features and AI Output.

“AI Features” means all generative artificial intelligence, machine learning, natural language processing, large language model, computer vision, automated data enrichment functionality or similar feature used to provide the Services whether or not designated as “AI” in the Documentation.

“AI Output(s)” means content, data, predictions, risk scores, recommendations, property attribute fills, data quality assessments, enriched statements of values, or other results generated by AI Features.

“Model Provider” means any third-party provider of artificial intelligence models, model hosting, inference, or related services utilized by Archipelago in providing AI Features (e.g., Anthropic).

2.2 AI Features. The Service includes the use of AI Features to produce AI Outputs. Archipelago will utilize Model Providers and ensure that they are subject to data handling obligations consistent with Archipelago’s obligations under this Agreement. Archipelago will maintain a current list of Model Providers used in the Service, which is included in its list of subprocessors viewable at: [https://www.onarchipelago.com/customers/subprocessors].

2.3 AI Assets. Archipelago retains all right, title, and interest in and to the AI Assets and all improvements, modifications, and derivative works of the foregoing, whether developed independently or informed by processing Customer Data. Customer’s contribution of Customer Data to the Service does not create any ownership interest, license claim, or other right in favor of Customer or its customers or users in or to any AI Assets, notwithstanding that such AI Assets may have been influenced or improved by Customer Data.

2.4 Output Limitations. Customer acknowledges and agrees that AI Output is probabilistic and non-deterministic and may contain errors, inaccuracies, hallucinations, or biases. The quality and accuracy of Outputs may also be materially affected by the quality of Customer Data and input provided and the quality and accuracy of data received from third party data providers and public information sources.

2.5 AI Acceptable Use. In addition to the restrictions in Section 2.7 of the Agreement, Customer will not, and will not permit any third party to, use AI Features or Outputs to: (a) train, retrain, fine-tune, or develop any artificial intelligence or machine learning model that competes with Archipelago or its Model Providers; (b) represent any Outputs as being approved, vetted, or certified by Archipelago; (c) represent any Outputs as wholly human-generated; (d) generate harmful, illegal, or discriminatory content; (e) use AI Features in violation of any Model Provider acceptable use policy made available to Customer in the Documentation; or (f) use automated means to systematically extract Outputs from the Service at a scale beyond normal use, or redistribute, resell, or provide access to Outputs to third-party data vendors, competing services, or to any insurer or reinsurer for purposes of building competing data or analytics capabilities, without Archipelago’s prior written consent.

2.6 Human-in-the-Loop for Regulated Insurance Decisions. Customer agrees not to use Outputs (e.g., enriched property data, risk scores, data quality assessments) as the sole basis for insurance underwriting decisions, such as, without limitation: (i) binding, issuing, renewing, or canceling an insurance policy; (ii) setting or adjusting insurance premiums; (iii) denying, reducing, or adjudicating an insurance claim; or (iv) any other insurance decision constituting an “adverse action” under applicable law, without independent human review and validation by a qualified professional. Customer shall not use AI Features to make any decisions in any manner where such use could produce material financial, legal, or regulatory consequences or other similarly significant effect on any individual or entity, without appropriate human review and compliance with applicable law. Archipelago shall have no liability for any insurance or financial decision made by Customer or its Clients in reliance on Outputs without required human review. Where Customer grants any third party User access to Outputs through the Service, such access will be subject to and governed by the Archipelago User Terms. Customer shall not represent to any third party that Outputs have been independently verified or warranted to be accurate by Archipelago.

2.7 Platform as Data Management Tool. The Service (including the AI Features and Outputs) is provided as a data management tool that supports Customer's professional business activities; it does not constitute and shall not be positioned as insurance advisory, underwriting advisory, or actuarial advisory services. Archipelago is not a licensed insurance broker, agent, surplus lines broker, or insurance intermediary in any jurisdiction, and nothing in this Agreement or in Archipelago's provision of the Service constitutes the exercise of any licensed insurance function. Customer remains solely responsible for independently reviewing, validating, and determining the appropriateness of all Outputs before incorporating them into any carrier submission, coverage recommendation, or other insurance transaction. To the extent Archipelago's SOC 2 Type II certification, security documentation, or other similar documentation maintained by Archipelago is reasonably required by Customer for purposes of regulatory examination, vendor oversight, or insurance compliance requirements, Archipelago will make such documentation available upon reasonable written request subject to appropriate confidentiality protections. Additionally, if Customer is a broker: (i) Customer remains solely responsible, as a licensed professional, for taking professional responsibility for all Outputs before incorporating them into any insurance transaction on behalf of any underlying Client; (ii) Customer's professional standard of care with respect to its underlying Clients is not modified or limited by Archipelago's provision of the Service or by any Outputs generated thereunder.

2.8 Use of Model Providers. Customer acknowledges that Archipelago’s use of Model Providers is necessary and the provision of Services is contingent upon such use. Customer authorizes Archipelago to use Model Providers as necessary to provide the Services. Use of Model Providers does not reduce Archipelago’s responsibility for the Service and AI Features under this Agreement. Archipelago may temporarily suspend or limit AI Features or the Services to the extent required by the unavailability of a Model Provider, applicable law or a material security or safety risk. Archipelago will promptly notify Customer of any such suspension or limitation and will use commercially reasonable efforts to restore the affected AI Features. Any suspension under this Section does not limit the rights of either party under the Agreement. Customer is not required to enter into separate agreements with Model Providers.

2.9 No Training by Model Providers. Archipelago will not permit Model Providers to train, fine-tune or improve models using Customer Data. Archipelago will ensure its Model Providers are subject to contractual obligations regarding the processing of input and output that are materially consistent with the requirements of this Agreement. Notwithstanding the foregoing, ephemeral processing by Model Providers may occur. Such ephemeral processing is transient and generally does not occur beyond the time necessary to generate the applicable session or interaction. Model Providers may also retain and internally review input and output for a retention period (typically 30 days) solely for operational monitoring to detect, prevent and remediate abuse, misuse, security incidents, malfunctions or abnormal performance.

2.10 Internal Training by Archipelago. The Services include AI Features that develop over time by using data, including Customer Data. Notwithstanding anything to the contrary in this Agreement, Archipelago may collect and use Customer Data, along with Usage Data, Benchmark Data and Derived Data for purposes of training, re-training, and fine-tuning its internal proprietary models and developing, operating, supporting, improving Archipelago’s products and services, and as otherwise permitted under this Agreement. Archipelago shall not publish Usage Data or Derived Data except to the extent such data is aggregated and de-identified such that neither Customer nor Users are identified therein and such use is in accordance with the DPA.

2.11 Benchmark Features. Certain AI Features within the Service, such as data quality recommendations and valuation outlier analysis (“Benchmark Features”), operate by comparing Customer Data against statistical benchmarks derived from de-identified, aggregated data drawn from the portfolios of Archipelago’s customer base as a whole (“Benchmark Data”). Benchmark Features compare Customer Data against population-level statistical averages and do not expose identifiable data of any individual customer to any other customer. Customer’s use of the Service includes Customer’s consent to Archipelago incorporating de-identified, aggregated statistical patterns derived from Customer Data into the Benchmark Data pool used to improve Benchmark Features for all customers. Archipelago retains all rights in Benchmark Data and such rights survive termination of this Agreement. Archipelago has no obligation to remove Customer’s de-identified contribution from Benchmark Data following termination. If Customer is a broker, Customer’s consent under this Section extends to Customer Data submitted across all Client Portfolio Accounts under Customer’s subscription, subject to Customer’s obligations to make required disclosures to underlying Clients.

2.12 AI Regulation in Insurance. Customer acknowledges that use of Outputs in connection with insurance underwriting, pricing, claims handling, or coverage decisions may be subject to applicable state insurance laws, regulations and regulatory guidance governing the use of external data and algorithms in insurance, including without limitation, such laws as Colorado SB 21-169, the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, and similar state laws and regulations as enacted or amended from time to time as may be applicable to Customer, its Users, carriers, underwriters and other regulated participants in the insurance placement process. Customer is solely responsible for ensuring its use and its Users’ use of Outputs complies with all applicable insurance regulatory requirements and laws. Customer's obligation to notify third parties granted access to Outputs through the Service reflects in part the regulatory obligations applicable to those parties. Upon reasonable written request, Archipelago will provide reasonable information to assist Customer in meeting its regulatory compliance obligations, subject to Archipelago’s confidentiality and trade secret protections. Customer acknowledges that the quality and accuracy of Outputs may be materially affected by the quality of Customer Data provided, and Customer and its Users are solely responsible for evaluating Outputs before relying on it for any business or regulatory decision.

2.13 EU AI Act Disclosure. The EU Artificial Intelligence Act ("EU AI Act") may apply to AI Features where Customer uses the Service in the European Union. The classification of AI systems used in connection with insurance underwriting, placement, or risk assessment under the EU AI Act is subject to ongoing regulatory guidance from the European Commission and has not been finally determined. Archipelago will maintain technical documentation regarding its AI Features in a manner designed to support Customer's compliance assessments under the EU AI Act. Customer is solely responsible for assessing whether its specific use of AI Features and AI Output in its jurisdiction constitutes High-Risk AI use under applicable EU law and for satisfying any resulting compliance obligations, including obligations relating to transparency, human oversight, and record-keeping. Archipelago will cooperate with Customer's reasonable requests for information to support such assessment, subject to Archipelago's confidentiality and trade secret protections.

2.14 Third-Party Integration Liability. Where the Service facilitates data exchange between Customer and third-party catastrophe modeling tools, risk management information systems, or other insurance industry platforms (“Integrated Systems”), Archipelago is not responsible for the accuracy, availability, or outputs of such Integrated Systems, nor for any losses arising from data transformation or transmission errors occurring within or between Integrated Systems outside of Archipelago’s direct control. Customer is responsible for validating data transmitted to or received from Integrated Systems through the Service.

2.15 AI Disclaimer. WITHOUT LIMITING THE DISCLAIMER IN SECTION 9.4 OF THE AGREEMENT, ARCHIPELAGO EXPRESSLY DISCLAIMS ANY WARRANTY THAT OUTPUTS, INCLUDING ENRICHED PROPERTY DATA, RISK SCORES, DATA QUALITY ASSESSMENTS, OR RECOMMENDED DATA CORRECTIONS, ARE ACCURATE, COMPLETE, UNIQUE, OR SUITABLE FOR USE IN INSURANCE UNDERWRITING, PREMIUM CALCULATION, CATASTROPHE MODELING, CLAIMS HANDLING, OR ANY OTHER INSURANCE TRANSACTION. AI OUTPUT IS PROBABILISTIC AND NON-DETERMINISTIC AND MAY CONTAIN ERRORS, INACCURACIES, HALLUCINATIONS, OR BIASES. DIFFERENT CUSTOMERS MAY RECEIVE IDENTICAL OR SIMILAR OUTPUTS FOR SIMILAR INPUTS. ARCHIPELAGO IS NOT A LICENSED INSURANCE ADVISOR, ACTUARY, OR RISK CONSULTANT, AND OUTPUTS DO NOT CONSTITUTE PROFESSIONAL INSURANCE, ACTUARIAL, OR RISK MANAGEMENT ADVICE. CUSTOMER IS SOLELY RESPONSIBLE FOR INDEPENDENTLY VERIFYING OUTPUTS BEFORE INCORPORATING IT INTO ANY INSURANCE TRANSACTION, REGULATORY FILING, OR FINANCIAL DECISION.

2.16 AI Indemnity. Archipelago’s indemnification obligations under this Section 11.1 of the Agreement also extend to third-party claims alleging that Customer’s authorized use of Outputs infringes a third-party copyright (an “Output Claim”), subject to the following additional exclusions: Output Claims are excluded to the extent resulting from (i) Customer’s failure to use AI Features in accordance with the Documentation, including failure to use provided safety filters or guardrails; (ii) Customer providing infringing inputs, prompts and instructions (“Input”) or Input designed to, or reasonably likely to, generate infringing Output; or (iii) Customer modifying Output or combining it with other materials where the alleged infringement would not have occurred but for such modification or combination. Output Claims are subject solely to the General Liability Cap, regardless of any higher caps that may otherwise apply. Customer’s indemnification obligations under Section 11.2 of the Agreement shall extend to Claims that Customer’s Input caused the Service to generate Outputs that infringe third-party intellectual property rights, violate applicable law, or violate applicable insurance regulations.