| |
AI Acceptable Use Policy for Platform Services (Medical AI) | |
Version: v1.0 June 2026
ARTIFICIAL INTELLIGENCE ACCEPTABLE USE POLICY
1. INTRODUCTION
1.1. This AI AUP:
1.1.1. forms part of the Agreement (as described in the Platform GTCs);
1.1.2. has been prepared by the Supplier for the purpose of information and technical transparency (including for the purposes of the AI Act); and
1.1.3. may be updated and amended by the Supplier from time to time and at its sole discretion.
1.2. In this AI AUP defined words and phrases shall have their respective meanings given in the Platform GTCs.
2.1. The Supplier provides the following transparency information in respect of the AI elements of the Platform Service:
2.1.1. Technical nature of the AI System. The AI Features comprise:
2.1.1.1. optical character recognition (OCR) (which is applied to scanned and image-based medical records);
2.1.1.2. document classification, de-duplication, chronological ordering and pagination logic; and
2.1.1.3. AI Models for: information extraction; indexation; and generating medical chronologies and summaries (including to establish and present facts).
The Features operate on a per-instruction basis meaning that Customer’s Data is not used to train, fine-tune or otherwise improve the underlying model for the benefit of AI Model’s provider. Anonymised Prompts and Outputs may be used by the Supplier but only to the extent described in paragraph 8 below.
2.1.2. Third party technologies and hosted services. The AI System is powered by the following third-party sub-processors:
2.1.2.1. AI Model and content extraction provider(s): Amazon Bedrock, as a wrapper for AI Models and other technologies, and other text extraction capabilities that are operated within the Supplier’s virtual private connection (which utilises the AWS private link). Amazon Bedrock is operated under zero-retention terms and is not trained on the Customer’s Data.
2.1.2.2. Hosting and storage: AWS, region: eu-west-2 (London).
2.1.2.3. Third Party Access to Prompts and Outputs. Sub-processors receive inputs and system Prompts solely to return the requested Output. They do not retain, log beyond transient processing, or use the data for any other purpose, including model training.
2.1.3. Purpose of the AI Features and intended Outputs. The AI Features exist to compress the manual time required to organise and review medical records. The intended Outputs are as follows:
2.1.3.1. indexed, paginated and chronologically ordered medical record bundles;
2.1.3.2. structured medical chronologies and summaries identifying key events, treatments, investigations and clinician entries; and other content; and
2.1.3.3. search and extraction functions over the compiled bundle.
The Outputs are administrative and assistive in nature. They are produced to support — not replace — the legal and clinical judgement of the qualified fee-earner or expert instructed on the matter.
2.1.4. Limitations of the AI Features.
IMPORTANT: The Customer and the Authorised Users must note the following:
2.1.4.1. Outputs are probabilistic and may contain errors, omissions or mis-attributions, including occasional fabricated content (so-called "hallucination");
2.1.4.2. the accuracy of Outputs is materially dependent on the legibility and completeness of the source records; poor-quality scans, handwritten notes and non-standard formats reduce extraction accuracy, although confidence scores, and other user feedback is provided where the AI System has been able to identify areas requiring the Authorised User’s consideration;
2.1.4.3. the AI System is not designed or warranted for clinical diagnosis, treatment decisions, prognostic assessment, quantum calculation or the provision of legal advice;
2.1.4.4. Outputs must be reviewed carefully and verified by a suitably qualified person before being relied on for any substantive decision, pleading, report or correspondence;
2.1.4.5. the AI System is not designed to identify every relevant entry in a record set; Authorised Users should treat Outputs as a working draft to be checked against source documents.
2.1.5. Monitoring, logging and error mitigation. The Supplier maintains the following controls:
2.1.5.1. Access logging: all internal access to Customer’s Data is logged and reviewable on request; and
2.1.5.2. AI Model and Prompt version control: the AI Model version, prompt version and processing parameters used to generate each Output are recorded to support the tracing and reproduction of Outputs to the extent reasonably and technically possible.
3. HUMAN OVERSIGHT
3.1. The Customer warrants and undertakes to the Supplier that there will be human oversight of the use of the Platform Services at all times in which it is in use.
3.2. The Customer acknowledges paragraph 2.1.4 (Limitations of the AI Features) above and:
3.2.1. shall: (i) not use the Outputs as the sole basis for any decision that produces legal or similarly significant effects on a data subject (including decisions concerning settlement, quantum, causation, breach of duty, prognosis, treatment, or eligibility for a claim). Any such decision must be made by a suitably qualified person exercising independent professional judgement; (ii) ensure that all Authorised Users are appropriately qualified, trained on the limitations of the AI System as set out in this AI AUP, and instructed on the review obligations in this AI AUP; (iii) maintain its own records of review and verification sufficient to demonstrate compliance with this AI AUP and with its professional, regulatory and legal obligations; (iv) not represent any Output as the unaided work product of a human author; (v) report any material errors, omissions or anomalies in the Outputs to the Supplier; and (vi) comply with all applicable professional conduct rules (including, where relevant, SRA Standards and Regulations, Bar Standards Board Handbook, GMC Good Medical Practice, or equivalent) in connection with its use of the Outputs; and
3.2.2. is responsible for implementing review procedures proportionate to the use to which the Outputs are put and the potential consequences of error.
3.3. Nothing in the Agreement (or through use of the Platform Services) is intended to, and the Customer shall not treat it as, a delegation of any reserved legal activity, regulated medical activity, or other professional function that must be performed by a qualified person under applicable law or regulatory rules.
4.1. The Customer acknowledges and agrees that:
4.1.1. paragraphs 2.1.4 (Limitations of the AI Features) and 5 (Limitations and Disclaimers) apply to all use of the AI Systems deployed to the Platform Services and Outputs;
4.1.2. the AI Systems used in the Platform Services:
4.1.2.1. are under continuous development; and
4.1.2.2. use Generative AI to create Outputs;
4.1.3. the AI Systems deployed to the Platform Services and Outputs are not intended to be used for the following purposes:
4.1.3.1. Clinical or treatment purposes. The Platform Service is intended solely for use in connection with the preparation, investigation, prosecution, defence or resolution of legal claims, matter, and related medico-legal purposes set out in the Order. The Customer shall not use the Platform Service or any Output for: (i) clinical diagnosis, differential diagnosis, or the exclusion of any condition; (ii) any decision concerning the treatment, management, medication, referral, monitoring or discharge of a patient; (iii) triage, screening or prioritisation of patients for clinical care; (iv) prognostic assessment used to inform clinical decision-making; (v) the provision of medical advice to any individual; or (vi) any other purpose that would constitute the practice of medicine or the provision of healthcare.
4.1.3.2. Regulated medical device use. The Platform Service is not a medical device and has not been developed, tested, validated, certified or approved as a medical device under the UK Medical Devices Regulations 2002, EU Regulation (EU) 2017/745 (MDR), the US Federal Food, Drug, and Cosmetic Act, or any equivalent regulatory regime. The Customer shall not use, hold out, market, or deploy the Platform Service or any Output as a medical device, clinical decision support tool, or in vitro diagnostic.
4.1.3.3. Safety-critical use. The Customer shall not use the Platform Service or any Output in any setting where failure, error or inaccuracy could reasonably be expected to result in death, personal injury, deterioration of health, or serious environmental or property damage.
4.1.3.4. Reserved legal activities and unauthorised practice. The Customer shall not use the Platform Service or any Output: (i) to perform any reserved legal activity (within the meaning of section 12 of the Legal Services Act 2007) other than by a person authorised to do so; (ii) as a substitute for the exercise of independent professional legal judgement by a qualified fee-earner; or (iii) in any manner that would breach applicable professional conduct rules, including the SRA Standards and Regulations, the Bar Standards Board Handbook, CILEX Code of Conduct, or equivalent rules in any other jurisdiction.
4.1.3.5. Automated decision-making with legal or similarly significant effect. The Customer shall not use any Output as the sole basis for a decision producing legal or similarly significant effects on any data subject within the meaning of Article 22 UK GDPR, including decisions on liability, causation, quantum, settlement, eligibility, or the acceptance or rejection of a claim.
4.1.3.6. Use outside the agreed data scope. The Customer shall not submit to the Platform Service any data other than medical records and associated medico-legal documents reasonably required for the preparation of the relevant claim or matter or medico legal purpose set out in the Order. Without limitation, the Customer shall not submit: (i) records relating to individuals who are not claimants, defendants, witnesses or otherwise relevant parties to a legal matter being handled by the Customer; (ii) special category data unrelated to the medico-legal matter; (iii) classified, military, intelligence or national security information; or (iv) data the Customer is not lawfully entitled to process.
4.1.3.7. Re-identification and unlawful processing. The Customer shall not use the Platform Service or any Output to: (i) re-identify any individual from de-identified or pseudonymised data; (ii) profile or survey any individual for purposes unrelated to the medico-legal matter; (iii) carry out any processing in breach of UK GDPR, the Data Protection Act 2018, or equivalent data protection legislation; or (iv) process special category data otherwise than in reliance on a valid Article 9 condition and in accordance with the DPA.
4.1.3.8. Training of third party AI models. The Customer shall not use the Platform Service or any Output to train, fine-tune, evaluate, benchmark or otherwise develop any artificial intelligence or machine learning model, other than internal models used solely for the Customer's own quality assurance in connection with the medico-legal matters for which the Platform Service is licensed.
4.1.3.9. Misrepresentation of Outputs. The Customer shall not: (i) present any Output as the unaided work product of a human author; (ii) represent any Output as having been verified, reviewed or endorsed by the Supplier beyond the warranties expressly given in this Agreement; or (iii) use any Output in a manner that misleads any court, tribunal, regulator, expert, opposing party, or client as to its origin or limitations.
4.1.3.10. Prohibited use generally. The Customer shall not use the Platform Service or any Output for any purpose that is prohibited by the Platform GTCs.
4.1.4. the AI Systems comprise and/or power an automated service that is:
4.1.4.1. intended to perform narrow procedural tasks;
4.1.4.2. designed for the Purpose;
4.1.5. it shall not (and shall ensure that all Authorised Users do not) use the AI Systems or Outputs for any purpose other than the Purpose and in compliance with the Agreement (including this AI AUP); and
4.1.6. further information concerning the technology (including third party infrastructure, systems, and technologies) powering the AI Systems is described in paragraph 2 (Transparency Information).
5. LIMITATIONS AND DISCLAIMERS
5.1. Given paragraph 4 (Acknowledgements) and without limiting paragraph 2.1.4 (Limitations of the AI Features), the Customer unconditionally and irrevocably agrees that:
5.1.1. the AI Systems and the Outputs are made available by the Supplier strictly for the Purpose;
5.1.2. given the nature of all Generative AI and the matters set out in paragraph 6 (Prompts), the Supplier cannot, and does not, warrant, undertake, covenant, or represent that Outputs are complete, accurate, or fit for the Customer’s or an Authorised User’s specific requirements;
5.1.3. the Customer assumes sole responsibility for its reliance on Outputs obtained through its interaction with the Platform Services and the conclusions or courses-of-action drawn from such use and is required to review all Outputs before use (including in litigation);
5.1.4. it must use the Prompts and Outputs in compliance with all applicable Legislation (including the guidance and mandatory requirements of:
5.1.4.1. the litigation venue (e.g., courts and tribunals and the CPR); and/or
5.1.4.2. any regulator having supervisory authority over its business and activities (e.g., the Solicitors’ Regulation Authority)); and
5.1.5. as a circuit breaker, the Supplier may limit and/or discontinue use and access to the Platform Service at its sole discretion and with or without giving prior notice (including by restricting the quantity, type and/or content of enquiries made to the AI Systems and Outputs drawn from use of the AI Systems).
5.2. Except if, and to the extent, binding applicable Legislation requires otherwise:
5.2.1. the Supplier makes the AI Systems and Outputs available without any warranties or guarantees of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, accuracy, completeness, reliability, security, or non-infringement; and
5.2.2. the Customer is solely responsible for any damages or liabilities arising from its (or an Authorised User’s) use of the AI Systems or Outputs.
6.1. The Customer:
6.1.1. is responsible for all Prompts submitted to the AI Systems by Authorised Users (including the quality, sufficiency, completeness, and accuracy of Prompts);
6.1.2. acknowledges and agrees that the quality and content of Outputs is dependent on the quality and content of the Prompts;
6.1.3. warrants, represents, undertakes, and covenants to the Supplier that it has all licences, permissions, and/or consents necessary to submit the data and/or information contained in the Prompts;
6.1.4. shall procure:
6.1.4.1. that each Authorised User is presented with, or informed of the existence and location of, this AI AUP; and
6.1.4.2. each Authorised User's compliance with this AI AUP;
6.2. Permitted scope of Prompts. The Customer shall ensure that all Prompts submitted to the AI Systems are limited to the legitimate medico-legal purpose for which the Platform Service is licensed, namely the review, indexing, summarisation, chronological organisation and factual interrogation of medical records and associated medico-legal documents in connection with a legal matter being handled by the Customer. Without limitation, permitted Prompts include those directed at:
6.2.1. identifying, extracting, summarising or chronologising entries within the medical records;
6.2.2. locating references to specific events, conditions, treatments, investigations, clinicians, dates or other factual matters appearing in the records;
6.2.3. cross-referencing entries within or between records in the bundle;
6.2.4. producing draft chronologies, summaries, indices or extracts derived from the records; and
6.2.5. other tasks ancillary to the preparation, investigation, prosecution, defence or resolution of a legal claim or matter or other medico-legal purpose set out in the Order.
6.3. Prohibited Prompts. The Customer shall not, and shall procure that no Authorised User shall, submit any Prompt that:
6.3.1. falls outside the medico-legal purpose — including Prompts that seek clinical advice, treatment recommendations, diagnoses, prognoses, or any other Output prohibited under this AI AUP;
6.3.2. manipulates or attempts to manipulate the AI System — including Prompts that attempt to:
6.3.2.1. override, bypass, disable or circumvent any safety control, content filter, system instruction, guardrail or operating parameter of the AI System;
6.3.2.2. cause the AI System to ignore, contradict or depart from the Supplier's system prompts, instructions or configuration;
6.3.2.3. elicit the disclosure of system prompts, model weights, configuration, sub-processor identities (beyond those disclosed under this AI AUP), or other confidential information of the Supplier or its sub-processors;
6.3.2.4. cause the AI System to impersonate a person, generate misleading attributions, or produce Outputs falsely presented as originating from a named clinician, expert, court or other third party; or
6.3.2.5. constitute a prompt injection, jailbreak, adversarial input, or similar technique intended to alter the intended behaviour of the AI System;
6.3.2.6. overload or degrade the Platform Service — including Prompts that:
6.3.2.6.1. are submitted at a volume, frequency or scale exceeding the Customer's fair use entitlement under this Agreement, or that are designed to consume disproportionate compute, storage or token capacity;
6.3.2.6.2. are generated by automated means, scripts, bots, or pipelines other than those expressly authorised by the Supplier;
6.3.2.6.3. constitute load testing, stress testing, denial-of-service or resource-exhaustion activity; or
6.3.2.6.4. are submitted with the intent or effect of impairing the availability, performance, security or integrity of the Platform Service for the Customer or any other customer of the Supplier;
6.3.2.7. are spurious or extraneous — including Prompts that:
6.3.2.7.1. have no reasonable connection to a live medico-legal matter being handled by the Customer;
6.3.2.7.2. are submitted for personal, recreational, experimental, exploratory or curiosity-driven purposes unrelated to the Customer's business;
6.3.2.7.3. are submitted to generate Outputs for use outside the Platform Service in a manner inconsistent with this Agreement;
6.3.2.7.4. seek to use the AI System as a general-purpose chatbot, search engine, drafting assistant for unrelated work, or for the production of content unrelated to medical record review and factual determination; or
6.3.2.7.5. are submitted to test, benchmark, evaluate or reverse-engineer the AI System other than as expressly permitted under this Agreement;
6.3.2.8. introduce prohibited content — including Prompts that:
6.3.2.8.1. contain malware, exploits, or executable code intended to compromise the Platform Service;
6.3.2.8.2. contain data of a category prohibited under paragraph 4.1.3.6, including data unrelated to the medico-legal matter, classified or military data, or data the Customer is not lawfully entitled to process; or
6.3.2.8.3. are unlawful, defamatory, harassing, discriminatory, or in breach of any third party right.
6.4. Monitoring and enforcement. The Customer acknowledges that the Supplier may, in accordance with this AI AUP and the DPA, monitor and log Prompts for the purposes of operating the Platform Service, enforcing this AI AUP, investigating suspected breaches, and responding to security or abuse incidents. The Supplier may, acting reasonably:
6.4.1. rate-limit, throttle or temporarily suspend Authorised Users or the Customer's access where Prompts breach this AI AUP;
6.4.2. refuse to process, or refuse to return Outputs in respect of, any Prompt that the Supplier reasonably considers to breach this AI AUP; and
6.4.3. notify the Customer of any material or repeated breach, with the Customer responsible for remediation within the Authorised User population.
6.5. Customer responsibility for Authorised Users. The Customer is responsible for the acts and omissions of its Authorised Users in connection with Prompts as if they were the acts and omissions of the Customer, and shall ensure that Authorised Users are trained on the permitted scope and prohibited categories set out in this AI AUP.
7. OUTPUTS
7.1. Given that the content, data, and/or information contained in the Outputs are created through the Supplier’s (and its licensors’) expertise and proprietary technologies, the Customer grants the Supplier a licence to use anonymised Outputs (including all intellectual property rights in the same) for the purposes set out in paragraph 8.
8. SUPPLIER’S USE OF ANONYMISED PROMPTS AND OUTPUTS
8.1. The Customer acknowledges and agrees that the Supplier may use effectively anonymised Prompts and/or Outputs (excluding reports created by Appointed Experts) for the following purposes:
8.1.1. improving the Platform and the Platform Services;
8.1.2. developing new or additional products and/or services; and
8.1.3. developing, training, and fine-tuning the Supplier’s AI Systems (including the Supplier’s virtual private connected deployment of the AI Models),
provided that, in doing so, it complies with all applicable Data Protection Legislation and Prompts and Outputs are not used to train the third party AI Model provider’s foundational AI Model.
8.2. For the purposes of paragraph 8.1, the phrase “effective anonymisation” means anonymising (through technical and organisational measures) the Prompt and/or Output such that it does not relate to an identified or identifiable natural person (i.e., the Prompt and/or Output is rendered anonymous in such a manner that the data subject is not or no longer identifiable or re-identifiable with a degree of certainty).