1. What this statement covers
LexLINQ is a legal technology and information platform for legal professionals and authorised business users. Its five focused modules are AI Matter Analysis, Legal Search, AI Chat, Document Compare, and Document Drafting.
LexLINQ assists defined tasks. It does not provide legal advice, represent any person, establish a solicitor–client relationship, or make final legal or professional decisions.
This statement describes our responsible-use approach and practical expectations. It is not a certification, accreditation, or representation that using LexLINQ automatically satisfies every law, professional duty, court requirement, client obligation, or organisational policy.
2. Accountability and meaningful human control
- A qualified legal professional must review and approve material before it is relied on, sent externally, signed, filed, or used for advice or a consequential decision.
- Users remain responsible for understanding client instructions, the complete factual record, current law, professional duties, and the likely consequences of using an output.
- Human review must be meaningful: the reviewer must have enough subject-matter knowledge, context, authority, and time to detect errors and reject or correct the result.
- AI output must not be presented as independently verified legal work unless that verification has actually occurred.
3. Appropriate and proportionate use
Use LexLINQ only where the task, data, reviewer capability, and available safeguards make AI assistance appropriate. The required controls should increase with the sensitivity of the information, complexity of the task, difficulty of detecting an error, and potential impact on a client or other person.
Do not use an AI workflow if the result cannot be adequately checked, the information cannot be submitted lawfully and ethically, or the applicable client, court, regulator, insurer, or organisational policy prohibits the use.
- Define the task and intended use before submitting information.
- Use the minimum information reasonably needed for that task.
- Confirm that the selected LexLINQ feature and its data scope match the intended purpose.
- Escalate uncertain, sensitive, novel, or high-impact work to an appropriately qualified person.
4. Reliability, limitations, and verification
Generative AI is probabilistic. An output may be incomplete, inaccurate, outdated, inconsistent, overly confident, or affected by ambiguous instructions and missing context. It may invent or misstate facts, citations, quotations, dates, legal tests, document contents, or calculations.
Fluent wording is not evidence of accuracy. A linked source helps verification but does not guarantee that an answer is exhaustive, current, correctly interpreted, or appropriate for the matter. Results may also change as models, providers, prompts, source collections, and product features change.
- Open and read every material cited authority and confirm the court, jurisdiction, currency, later treatment, pinpoint, and proposition.
- Compare factual statements, names, dates, quotations, obligations, and calculations against the original matter documents.
- Confirm that the completed, indexed matter documents are sufficiently complete for the task and investigate missing or contradictory material.
- Apply the organisation’s processes for precedent use, supervision, peer review, approval, quality assurance, and periodic evaluation.
5. Privacy, confidentiality, privilege, and security
Users must have authority to submit and process Customer Content. Before using client or third-party material, consider confidentiality, legal professional privilege, privacy, client instructions, contractual restrictions, court orders, permitted purposes, processing locations, retention, and provider access.
Minimise or de-identify personal, sensitive, privileged, or commercially valuable information where practical. Using a controlled AI service may reduce some risks compared with a public chatbot, but it does not remove professional duties or guarantee that privilege will survive an unauthorised or unnecessary disclosure.
Review the current Privacy Policy and Security page and request additional provider information where the client, matter, data category, or organisational policy requires a specific assessment.
6. Transparency, explainability, and records
People should not be misled about the role AI played in material work. Users must make any disclosure required by professional obligations, client arrangements, organisational policy, legislation, a court or tribunal, or the nature and impact of the use.
For material work, organisations should retain records appropriate to the risk, such as the feature and provider context, instructions, source material, output, verification performed, corrections, reviewer, approval, and any required disclosure. The record should be sufficient to explain the role of AI without retaining unnecessary confidential or personal information.
LexLINQ explanations and linked materials support review, but generative model reasoning is not fully observable and an explanation is not proof that an answer is correct.
7. Fairness, impact, and correction
AI outputs may reflect gaps, imbalance, stereotypes, or other unwanted bias in models, source material, prompts, or the surrounding workflow. Review whether assumptions, language, classifications, or recommendations could unfairly affect a person or group.
Do not use a LexLINQ output as the sole basis for a binding decision affecting a person’s rights, access, employment, credit, eligibility, or legal position. Affected people should have an appropriate path to human review, correction, or challenge where the context requires it.
- Check that relevant perspectives and contrary material have not been omitted.
- Distinguish verified facts from allegations, inferences, uncertainty, and generated content.
- Correct inaccurate personal information and material errors before further use.
- Escalate potentially discriminatory, harmful, or unexpectedly biased output.
8. AI providers and Customer Content
Depending on the feature and configuration, LexLINQ may send the minimum relevant content needed for a request to business or API services supplied by OpenAI, Google Gemini, or Anthropic. LexLINQ does not use Customer Content to train public AI models.
Providers, models, subprocessors, capabilities, data controls, and processing regions can change. A provider name does not by itself establish that a particular use is appropriate. Organisations should conduct and periodically revisit due diligence for their approved use cases.
9. Courts, tribunals, evidence, and formal documents
Before using AI-assisted material in a court, tribunal, transaction, advice, expert report, affidavit, witness statement, chronology, submission, or other formal document, identify and comply with the requirements applying to the particular forum, matter, document, and person.
These may include legislation, evidence rules, professional conduct duties, client obligations, court orders, practice directions, disclosure requirements, and requirements to verify facts, evidence, authorities, quotations, and chronologies. Rules differ by jurisdiction and change over time, so users must check the current position.
10. Prohibited use and system integrity
- Do not use LexLINQ to impersonate a lawyer, falsely claim entitlement to engage in legal practice, or misrepresent an AI output as legal advice from LexLINQ.
- Do not attempt to access another user or customer account, matter, document, prompt, output, credential, or protected system information.
- Do not use prompt injection, malicious documents, automated attacks, scraping, reverse engineering, or other techniques to bypass safeguards, access controls, or Usage Credit limits.
- Do not upload malware, unlawfully obtained material, or content that infringes rights or breaches confidentiality, privacy, privilege, a court order, or a contractual duty.
- Do not conceal material AI involvement where disclosure is required or present unchecked AI output as verified fact, law, evidence, expert opinion, or professional work.
11. Feedback, contestability, and continuing review
Report a harmful, inaccurate, biased, unsafe, or unexpected output, suspected data issue, or security concern to info@lexlinq.ai. Provide enough detail to investigate, but do not include unnecessary confidential, privileged, or personal information in ordinary email.
Feedback and incident reports support ongoing review of product behaviour, safeguards, user guidance, providers, and approved use cases. Users and organisations must also reassess their own use when features, models, law, professional guidance, court requirements, or the risk of a task changes.
