{"id":87444,"topic":"ai","source":"The National Law Review","title":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks - The National Law Review","url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","url_hash":"9ef0011cc081642aeaabaeafc247f1fb83c8939a","author":"","summary":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks The National Law Review","content":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.\nCourts are already ordering disclosure of AI prompts and outputs used in expert work, and Rule 29 stipulations that once shielded expert notes and communications are proving too narrow to cover AI chat logs. Companies, in-house counsel and litigation teams that retain experts need to understand how this exposure arises and what steps can limit the risk before an expert is confronted with a chat log at deposition or trial.\nThis Q&A examines the discovery risks created by expert witnesses’ use of generative AI and the practical steps litigation attorneys can take to manage them.\nHow Does Experts Witness AI Use Create Discovery Exposure?\nGenerative AI creates a category of record that did not meaningfully exist a few years ago, and every prompt an expert enters is a time-stamped record of what that expert was seeking, sometimes including the outcome the expert hoped to reach before the report was ever finalized. This transforms a process that was historically invisible into one that can be reconstructed in granular detail, giving opposing counsel a potential window into the expert’s reasoning, assumptions and even bias before the final report was written.\nLitigators are already incorporating requests for AI logs into fact and expert discovery, and those requests are on track to become standard practice as AI use spreads throughout the profession. Full session and prompt logs, system instructions given to the AI and any AI-assisted test exercises can all become discoverable material that shapes cross-examination.\nAre Courts Ordering Disclosure of AI Prompts and Logs?\nYes, and one federal court has already set a notable precedent. In Conservation Law Foundation Inc. v. Shell Oil Co., a US magistrate judge in the District of Connecticut ordered disclosure of the AI prompts and queries an expert used while producing her expert witness report.\nThe ruling matters because the parties had a Rule 29 discovery stipulation providing that there would be no discovery of each other’s expert notes, drafts or communications, and the court found that language too narrow to shield AI materials from disclosure. Parties relying on standard-form Rule 29 stipulations should not assume those provisions will protect AI conversations unless the stipulation specifically addresses AI use.\nWhat Should Expert Engagement Letters Say About AI Use?\nEngagement letters should be updated now, before an expert is retained and before any AI tool is used. Building AI-specific terms into the engagement letter gives counsel visibility into how an expert is working and creates a record that supports the credibility of the final opinion.\nEffective engagement letters typically address:\n- A requirement that the expert disclose which AI tools may be used during the engagement\n- An approval step requiring the expert to obtain counsel’s permission before using AI\n- Recordkeeping provisions ensuring prompts, iterative refinements and outputs are preserved and not overwritten\n- A verification requirement that the expert independently validate AI-generated content against the underlying facts and the expert’s own professional judgment\nThese provisions give counsel a defensible framework for explaining an expert’s process if AI use is later challenged, and they reduce the chance that an expert is caught unprepared during cross-examination.\nCan Parties Agree to Keep Expert AI Use Off Limits in Discovery?\nParties can attempt to stipulate under Rule 29 that neither side’s expert AI usage will be discoverable, and this remains a viable option worth raising early in a case. However, the Connecticut federal court’s ruling shows that general expert-materials stipulations will not automatically extend to AI logs, so any such agreement needs to specifically and unambiguously address AI prompts, outputs and related records.\nParties should also weigh the tradeoffs carefully before pursuing this route. Because AI use could substantially undermine an expert’s credibility, both sides may hesitate to fully close off this avenue of inquiry, particularly if either party suspects the other’s expert relied heavily on AI-generated content.\nAre Expert Witnesses Allowed to Use AI in Reports?\nGenerative AI is not inherently disqualifying for expert work, and there is a legitimate case for allowing measured use. Experts can use AI to play devil’s advocate, run mock cross-examinations or stress-test a report against an opposing expert’s likely positions, and used well, these tools can genuinely strengthen the rigor of an opinion.\nThe key is ensuring the expert can demonstrate that the final opinion reflects independent professional judgment rather than AI output. An expert should be prepared to show that AI-generated material was verified against the underlying facts, that the expert’s own expertise shaped the ultimate conclusions and that clear documentation connects the evidence to the opinion offered at trial.\nWhat Practical Steps Should Litigation Teams Take Now?\nCounsel retaining experts should assume that opposing counsel will request AI conversations and that those conversations, if left unmanaged, could seriously undermine the expert’s credibility at deposition or trial. Raising this issue with experts early and often, rather than waiting until AI use surfaces in discovery, is now a basic component of expert witness management.\nLitigation teams should build AI considerations into broader discovery strategy, both offensively and defensively, rather than treating AI use as an issue that only applies to their own experts. Requesting an opposing expert’s AI logs can be as valuable as protecting one’s own expert’s process, and teams that treat AI discovery as routine will be better positioned than those still treating it as a novel issue.\nHow Should Litigation Teams Build AI Into Their Discovery Strategy?\nGenerative AI has already moved from a theoretical discovery issue to a real source of trial exposure for expert witnesses. Courts are beginning to order disclosure of AI prompts and outputs, and existing Rule 29 stipulations covering expert notes and communications are proving insufficient to shield this new category of record.\nCompanies and counsel that update expert engagement letters now, with clear disclosure, approval, recordkeeping and verification requirements, will be far better positioned than those that wait until an AI chat log surfaces during a deposition. As generative AI becomes further embedded in expert practice, proactive governance of its use will increasingly separate credible expert testimony from testimony vulnerable to serious challenge.","image_url":"https://natlawreview.com/sites/default/files/styles/article_image/public/2026-09/AI%20Artificial%20Intelligence%20Phone%20Cloud%20Security-441581066_0.jpg.webp?itok=sJpYE2dz","lang":"en","published_at":"2026-09-21T18:57:37+00:00","fetched_at":"2026-09-21T20:15:05+00:00","status":"read","starred":0,"extract_state":"ok","summary_auto":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.","cluster_id":null,"extract_retries":0,"extract_error":null,"contract_version":"news_item.v1","format_contract_version":"news_item_formats.v1","dedup_url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 6987 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":6987,"summary_length":375,"usable_text_length":6987,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":6987,"summary_length":375}},"news_item":{"id":87444,"canonical_url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","source_url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","title":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks - The National Law Review","source_name":"The National Law Review","author":null,"published_at":"2026-09-21T18:57:37+00:00","locale":"en","topic":"ai","tags":[],"rss_summary":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks The National Law Review","full_text":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.\nCourts are already ordering disclosure of AI prompts and outputs used in expert work, and Rule 29 stipulations that once shielded expert notes and communications are proving too narrow to cover AI chat logs. Companies, in-house counsel and litigation teams that retain experts need to understand how this exposure arises and what steps can limit the risk before an expert is confronted with a chat log at deposition or trial.\nThis Q&A examines the discovery risks created by expert witnesses’ use of generative AI and the practical steps litigation attorneys can take to manage them.\nHow Does Experts Witness AI Use Create Discovery Exposure?\nGenerative AI creates a category of record that did not meaningfully exist a few years ago, and every prompt an expert enters is a time-stamped record of what that expert was seeking, sometimes including the outcome the expert hoped to reach before the report was ever finalized. This transforms a process that was historically invisible into one that can be reconstructed in granular detail, giving opposing counsel a potential window into the expert’s reasoning, assumptions and even bias before the final report was written.\nLitigators are already incorporating requests for AI logs into fact and expert discovery, and those requests are on track to become standard practice as AI use spreads throughout the profession. Full session and prompt logs, system instructions given to the AI and any AI-assisted test exercises can all become discoverable material that shapes cross-examination.\nAre Courts Ordering Disclosure of AI Prompts and Logs?\nYes, and one federal court has already set a notable precedent. In Conservation Law Foundation Inc. v. Shell Oil Co., a US magistrate judge in the District of Connecticut ordered disclosure of the AI prompts and queries an expert used while producing her expert witness report.\nThe ruling matters because the parties had a Rule 29 discovery stipulation providing that there would be no discovery of each other’s expert notes, drafts or communications, and the court found that language too narrow to shield AI materials from disclosure. Parties relying on standard-form Rule 29 stipulations should not assume those provisions will protect AI conversations unless the stipulation specifically addresses AI use.\nWhat Should Expert Engagement Letters Say About AI Use?\nEngagement letters should be updated now, before an expert is retained and before any AI tool is used. Building AI-specific terms into the engagement letter gives counsel visibility into how an expert is working and creates a record that supports the credibility of the final opinion.\nEffective engagement letters typically address:\n- A requirement that the expert disclose which AI tools may be used during the engagement\n- An approval step requiring the expert to obtain counsel’s permission before using AI\n- Recordkeeping provisions ensuring prompts, iterative refinements and outputs are preserved and not overwritten\n- A verification requirement that the expert independently validate AI-generated content against the underlying facts and the expert’s own professional judgment\nThese provisions give counsel a defensible framework for explaining an expert’s process if AI use is later challenged, and they reduce the chance that an expert is caught unprepared during cross-examination.\nCan Parties Agree to Keep Expert AI Use Off Limits in Discovery?\nParties can attempt to stipulate under Rule 29 that neither side’s expert AI usage will be discoverable, and this remains a viable option worth raising early in a case. However, the Connecticut federal court’s ruling shows that general expert-materials stipulations will not automatically extend to AI logs, so any such agreement needs to specifically and unambiguously address AI prompts, outputs and related records.\nParties should also weigh the tradeoffs carefully before pursuing this route. Because AI use could substantially undermine an expert’s credibility, both sides may hesitate to fully close off this avenue of inquiry, particularly if either party suspects the other’s expert relied heavily on AI-generated content.\nAre Expert Witnesses Allowed to Use AI in Reports?\nGenerative AI is not inherently disqualifying for expert work, and there is a legitimate case for allowing measured use. Experts can use AI to play devil’s advocate, run mock cross-examinations or stress-test a report against an opposing expert’s likely positions, and used well, these tools can genuinely strengthen the rigor of an opinion.\nThe key is ensuring the expert can demonstrate that the final opinion reflects independent professional judgment rather than AI output. An expert should be prepared to show that AI-generated material was verified against the underlying facts, that the expert’s own expertise shaped the ultimate conclusions and that clear documentation connects the evidence to the opinion offered at trial.\nWhat Practical Steps Should Litigation Teams Take Now?\nCounsel retaining experts should assume that opposing counsel will request AI conversations and that those conversations, if left unmanaged, could seriously undermine the expert’s credibility at deposition or trial. Raising this issue with experts early and often, rather than waiting until AI use surfaces in discovery, is now a basic component of expert witness management.\nLitigation teams should build AI considerations into broader discovery strategy, both offensively and defensively, rather than treating AI use as an issue that only applies to their own experts. Requesting an opposing expert’s AI logs can be as valuable as protecting one’s own expert’s process, and teams that treat AI discovery as routine will be better positioned than those still treating it as a novel issue.\nHow Should Litigation Teams Build AI Into Their Discovery Strategy?\nGenerative AI has already moved from a theoretical discovery issue to a real source of trial exposure for expert witnesses. Courts are beginning to order disclosure of AI prompts and outputs, and existing Rule 29 stipulations covering expert notes and communications are proving insufficient to shield this new category of record.\nCompanies and counsel that update expert engagement letters now, with clear disclosure, approval, recordkeeping and verification requirements, will be far better positioned than those that wait until an AI chat log surfaces during a deposition. As generative AI becomes further embedded in expert practice, proactive governance of its use will increasingly separate credible expert testimony from testimony vulnerable to serious challenge.","excerpt":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.","extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 6987 characters.","diagnostics_url":"/api/diagnose?url=https%3A//natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 6987 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":6987,"summary_length":375,"usable_text_length":6987,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":6987,"summary_length":375}}},"display_formats":["compact","card","full","digest_section","json"]},"daily_stack_record":{"title":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks - The National Law Review","url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","summary":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.","source":"The National Law Review","date":"2026-09-21T18:57:37+00:00","content":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.\nCourts are already ordering disclosure of AI prompts and outputs used in expert work, and Rule 29 stipulations that once shielded expert notes and communications are proving too narrow to cover AI chat logs. Companies, in-house counsel and litigation teams that retain experts need to understand how this exposure arises and what steps can limit the risk before an expert is confronted with a chat log at deposition or trial.\nThis Q&A examines the discovery risks created by expert witnesses’ use of generative AI and the practical steps litigation attorneys can take to manage them.\nHow Does Experts Witness AI Use Create Discovery Exposure?\nGenerative AI creates a category of record that did not meaningfully exist a few years ago, and every prompt an expert enters is a time-stamped record of what that expert was seeking, sometimes including the outcome the expert hoped to reach before the report was ever finalized. This transforms a process that was historically invisible into one that can be reconstructed in granular detail, giving opposing counsel a potential window into the expert’s reasoning, assumptions and even bias before the final report was written.\nLitigators are already incorporating requests for AI logs into fact and expert discovery, and those requests are on track to become standard practice as AI use spreads throughout the profession. Full session and prompt logs, system instructions given to the AI and any AI-assisted test exercises can all become discoverable material that shapes cross-examination.\nAre Courts Ordering Disclosure of AI Prompts and Logs?\nYes, and one federal court has already set a notable precedent. In Conservation Law Foundation Inc. v. Shell Oil Co., a US magistrate judge in the District of Connecticut ordered disclosure of the AI prompts and queries an expert used while producing her expert witness report.\nThe ruling matters because the parties had a Rule 29 discovery stipulation providing that there would be no discovery of each other’s expert notes, drafts or communications, and the court found that language too narrow to shield AI materials from disclosure. Parties relying on standard-form Rule 29 stipulations should not assume those provisions will protect AI conversations unless the stipulation specifically addresses AI use.\nWhat Should Expert Engagement Letters Say About AI Use?\nEngagement letters should be updated now, before an expert is retained and before any AI tool is used. Building AI-specific terms into the engagement letter gives counsel visibility into how an expert is working and creates a record that supports the credibility of the final opinion.\nEffective engagement letters typically address:\n- A requirement that the expert disclose which AI tools may be used during the engagement\n- An approval step requiring the expert to obtain counsel’s permission before using AI\n- Recordkeeping provisions ensuring prompts, iterative refinements and outputs are preserved and not overwritten\n- A verification requirement that the expert independently validate AI-generated content against the underlying facts and the expert’s own professional judgment\nThese provisions give counsel a defensible framework for explaining an expert’s process if AI use is later challenged, and they reduce the chance that an expert is caught unprepared during cross-examination.\nCan Parties Agree to Keep Expert AI Use Off Limits in Discovery?\nParties can attempt to stipulate under Rule 29 that neither side’s expert AI usage will be discoverable, and this remains a viable option worth raising early in a case. However, the Connecticut federal court’s ruling shows that general expert-materials stipulations will not automatically extend to AI logs, so any such agreement needs to specifically and unambiguously address AI prompts, outputs and related records.\nParties should also weigh the tradeoffs carefully before pursuing this route. Because AI use could substantially undermine an expert’s credibility, both sides may hesitate to fully close off this avenue of inquiry, particularly if either party suspects the other’s expert relied heavily on AI-generated content.\nAre Expert Witnesses Allowed to Use AI in Reports?\nGenerative AI is not inherently disqualifying for expert work, and there is a legitimate case for allowing measured use. Experts can use AI to play devil’s advocate, run mock cross-examinations or stress-test a report against an opposing expert’s likely positions, and used well, these tools can genuinely strengthen the rigor of an opinion.\nThe key is ensuring the expert can demonstrate that the final opinion reflects independent professional judgment rather than AI output. An expert should be prepared to show that AI-generated material was verified against the underlying facts, that the expert’s own expertise shaped the ultimate conclusions and that clear documentation connects the evidence to the opinion offered at trial.\nWhat Practical Steps Should Litigation Teams Take Now?\nCounsel retaining experts should assume that opposing counsel will request AI conversations and that those conversations, if left unmanaged, could seriously undermine the expert’s credibility at deposition or trial. Raising this issue with experts early and often, rather than waiting until AI use surfaces in discovery, is now a basic component of expert witness management.\nLitigation teams should build AI considerations into broader discovery strategy, both offensively and defensively, rather than treating AI use as an issue that only applies to their own experts. Requesting an opposing expert’s AI logs can be as valuable as protecting one’s own expert’s process, and teams that treat AI discovery as routine will be better positioned than those still treating it as a novel issue.\nHow Should Litigation Teams Build AI Into Their Discovery Strategy?\nGenerative AI has already moved from a theoretical discovery issue to a real source of trial exposure for expert witnesses. Courts are beginning to order disclosure of AI prompts and outputs, and existing Rule 29 stipulations covering expert notes and communications are proving insufficient to shield this new category of record.\nCompanies and counsel that update expert engagement letters now, with clear disclosure, approval, recordkeeping and verification requirements, will be far better positioned than those that wait until an AI chat log surfaces during a deposition. As generative AI becomes further embedded in expert practice, proactive governance of its use will increasingly separate credible expert testimony from testimony vulnerable to serious challenge.","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 6987 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 6987 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":6987,"summary_length":375,"usable_text_length":6987,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":6987,"summary_length":375}},"tags":[]},"fallback_formats":["markdown","json","html"],"actions":{"read":"/item/87444","export_markdown":"/api/items/87444/export?format=markdown","export_json":"/api/items/87444/export?format=json","diagnose":"/api/diagnose?url=https%3A//natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks"},"formats":{"full":{"id":87444,"title":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks - The National Law Review","url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","source":"The National Law Review","author":null,"published_at":"2026-09-21T18:57:37+00:00","locale":"en","topic":"ai","tags":[],"excerpt":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.","full_text":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.\nCourts are already ordering disclosure of AI prompts and outputs used in expert work, and Rule 29 stipulations that once shielded expert notes and communications are proving too narrow to cover AI chat logs. Companies, in-house counsel and litigation teams that retain experts need to understand how this exposure arises and what steps can limit the risk before an expert is confronted with a chat log at deposition or trial.\nThis Q&A examines the discovery risks created by expert witnesses’ use of generative AI and the practical steps litigation attorneys can take to manage them.\nHow Does Experts Witness AI Use Create Discovery Exposure?\nGenerative AI creates a category of record that did not meaningfully exist a few years ago, and every prompt an expert enters is a time-stamped record of what that expert was seeking, sometimes including the outcome the expert hoped to reach before the report was ever finalized. This transforms a process that was historically invisible into one that can be reconstructed in granular detail, giving opposing counsel a potential window into the expert’s reasoning, assumptions and even bias before the final report was written.\nLitigators are already incorporating requests for AI logs into fact and expert discovery, and those requests are on track to become standard practice as AI use spreads throughout the profession. Full session and prompt logs, system instructions given to the AI and any AI-assisted test exercises can all become discoverable material that shapes cross-examination.\nAre Courts Ordering Disclosure of AI Prompts and Logs?\nYes, and one federal court has already set a notable precedent. In Conservation Law Foundation Inc. v. Shell Oil Co., a US magistrate judge in the District of Connecticut ordered disclosure of the AI prompts and queries an expert used while producing her expert witness report.\nThe ruling matters because the parties had a Rule 29 discovery stipulation providing that there would be no discovery of each other’s expert notes, drafts or communications, and the court found that language too narrow to shield AI materials from disclosure. Parties relying on standard-form Rule 29 stipulations should not assume those provisions will protect AI conversations unless the stipulation specifically addresses AI use.\nWhat Should Expert Engagement Letters Say About AI Use?\nEngagement letters should be updated now, before an expert is retained and before any AI tool is used. Building AI-specific terms into the engagement letter gives counsel visibility into how an expert is working and creates a record that supports the credibility of the final opinion.\nEffective engagement letters typically address:\n- A requirement that the expert disclose which AI tools may be used during the engagement\n- An approval step requiring the expert to obtain counsel’s permission before using AI\n- Recordkeeping provisions ensuring prompts, iterative refinements and outputs are preserved and not overwritten\n- A verification requirement that the expert independently validate AI-generated content against the underlying facts and the expert’s own professional judgment\nThese provisions give counsel a defensible framework for explaining an expert’s process if AI use is later challenged, and they reduce the chance that an expert is caught unprepared during cross-examination.\nCan Parties Agree to Keep Expert AI Use Off Limits in Discovery?\nParties can attempt to stipulate under Rule 29 that neither side’s expert AI usage will be discoverable, and this remains a viable option worth raising early in a case. However, the Connecticut federal court’s ruling shows that general expert-materials stipulations will not automatically extend to AI logs, so any such agreement needs to specifically and unambiguously address AI prompts, outputs and related records.\nParties should also weigh the tradeoffs carefully before pursuing this route. Because AI use could substantially undermine an expert’s credibility, both sides may hesitate to fully close off this avenue of inquiry, particularly if either party suspects the other’s expert relied heavily on AI-generated content.\nAre Expert Witnesses Allowed to Use AI in Reports?\nGenerative AI is not inherently disqualifying for expert work, and there is a legitimate case for allowing measured use. Experts can use AI to play devil’s advocate, run mock cross-examinations or stress-test a report against an opposing expert’s likely positions, and used well, these tools can genuinely strengthen the rigor of an opinion.\nThe key is ensuring the expert can demonstrate that the final opinion reflects independent professional judgment rather than AI output. An expert should be prepared to show that AI-generated material was verified against the underlying facts, that the expert’s own expertise shaped the ultimate conclusions and that clear documentation connects the evidence to the opinion offered at trial.\nWhat Practical Steps Should Litigation Teams Take Now?\nCounsel retaining experts should assume that opposing counsel will request AI conversations and that those conversations, if left unmanaged, could seriously undermine the expert’s credibility at deposition or trial. Raising this issue with experts early and often, rather than waiting until AI use surfaces in discovery, is now a basic component of expert witness management.\nLitigation teams should build AI considerations into broader discovery strategy, both offensively and defensively, rather than treating AI use as an issue that only applies to their own experts. Requesting an opposing expert’s AI logs can be as valuable as protecting one’s own expert’s process, and teams that treat AI discovery as routine will be better positioned than those still treating it as a novel issue.\nHow Should Litigation Teams Build AI Into Their Discovery Strategy?\nGenerative AI has already moved from a theoretical discovery issue to a real source of trial exposure for expert witnesses. Courts are beginning to order disclosure of AI prompts and outputs, and existing Rule 29 stipulations covering expert notes and communications are proving insufficient to shield this new category of record.\nCompanies and counsel that update expert engagement letters now, with clear disclosure, approval, recordkeeping and verification requirements, will be far better positioned than those that wait until an AI chat log surfaces during a deposition. As generative AI becomes further embedded in expert practice, proactive governance of its use will increasingly separate credible expert testimony from testimony vulnerable to serious challenge.","reading_time_min":5,"extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 6987 characters.","diagnostics_url":"/api/diagnose?url=https%3A//natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 6987 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":6987,"summary_length":375,"usable_text_length":6987,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":6987,"summary_length":375}}},"quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 6987 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":6987,"summary_length":375,"usable_text_length":6987,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":6987,"summary_length":375}},"actions":{"read":"/item/87444","export_markdown":"/api/items/87444/export?format=markdown","export_json":"/api/items/87444/export?format=json","diagnose":"/api/diagnose?url=https%3A//natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks"}},"digest":{"id":87444,"title":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks - The National Law Review","url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","source":"The National Law Review","topic":"ai","published_at":"2026-09-21T18:57:37+00:00","excerpt":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their…","quality_bucket":"high","quality_reason":"High confidence: full text extraction produced 6987 characters.","reading_time_min":5,"cluster_id":null},"card":{"display_title":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks - The National Law Review","subtitle":"The National Law Review · 2026-09-21","summary":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports…","badges":["quality:high"],"links":{"read":"/item/87444","original":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","diagnose":"/api/diagnose?url=https%3A//natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks"},"quality_warning":null},"export":{"title":"AI in Expert Witness Testimony- What Litigators Need to Know About LLM Discovery Risks - The National Law Review","url":"https://natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","summary":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.","source":"The National Law Review","date":"2026-09-21T18:57:37+00:00","content":"Generative AI tools have quietly entered the expert witness process, and recent litigation developments show how exposed that process can become. Experts who use AI tools like ChatGPT or other LLMs to help draft reports may be creating a detailed, time-stamped record of their reasoning that opposing counsel can obtain and use to undermine the expert at deposition or trial.\nCourts are already ordering disclosure of AI prompts and outputs used in expert work, and Rule 29 stipulations that once shielded expert notes and communications are proving too narrow to cover AI chat logs. Companies, in-house counsel and litigation teams that retain experts need to understand how this exposure arises and what steps can limit the risk before an expert is confronted with a chat log at deposition or trial.\nThis Q&A examines the discovery risks created by expert witnesses’ use of generative AI and the practical steps litigation attorneys can take to manage them.\nHow Does Experts Witness AI Use Create Discovery Exposure?\nGenerative AI creates a category of record that did not meaningfully exist a few years ago, and every prompt an expert enters is a time-stamped record of what that expert was seeking, sometimes including the outcome the expert hoped to reach before the report was ever finalized. This transforms a process that was historically invisible into one that can be reconstructed in granular detail, giving opposing counsel a potential window into the expert’s reasoning, assumptions and even bias before the final report was written.\nLitigators are already incorporating requests for AI logs into fact and expert discovery, and those requests are on track to become standard practice as AI use spreads throughout the profession. Full session and prompt logs, system instructions given to the AI and any AI-assisted test exercises can all become discoverable material that shapes cross-examination.\nAre Courts Ordering Disclosure of AI Prompts and Logs?\nYes, and one federal court has already set a notable precedent. In Conservation Law Foundation Inc. v. Shell Oil Co., a US magistrate judge in the District of Connecticut ordered disclosure of the AI prompts and queries an expert used while producing her expert witness report.\nThe ruling matters because the parties had a Rule 29 discovery stipulation providing that there would be no discovery of each other’s expert notes, drafts or communications, and the court found that language too narrow to shield AI materials from disclosure. Parties relying on standard-form Rule 29 stipulations should not assume those provisions will protect AI conversations unless the stipulation specifically addresses AI use.\nWhat Should Expert Engagement Letters Say About AI Use?\nEngagement letters should be updated now, before an expert is retained and before any AI tool is used. Building AI-specific terms into the engagement letter gives counsel visibility into how an expert is working and creates a record that supports the credibility of the final opinion.\nEffective engagement letters typically address:\n- A requirement that the expert disclose which AI tools may be used during the engagement\n- An approval step requiring the expert to obtain counsel’s permission before using AI\n- Recordkeeping provisions ensuring prompts, iterative refinements and outputs are preserved and not overwritten\n- A verification requirement that the expert independently validate AI-generated content against the underlying facts and the expert’s own professional judgment\nThese provisions give counsel a defensible framework for explaining an expert’s process if AI use is later challenged, and they reduce the chance that an expert is caught unprepared during cross-examination.\nCan Parties Agree to Keep Expert AI Use Off Limits in Discovery?\nParties can attempt to stipulate under Rule 29 that neither side’s expert AI usage will be discoverable, and this remains a viable option worth raising early in a case. However, the Connecticut federal court’s ruling shows that general expert-materials stipulations will not automatically extend to AI logs, so any such agreement needs to specifically and unambiguously address AI prompts, outputs and related records.\nParties should also weigh the tradeoffs carefully before pursuing this route. Because AI use could substantially undermine an expert’s credibility, both sides may hesitate to fully close off this avenue of inquiry, particularly if either party suspects the other’s expert relied heavily on AI-generated content.\nAre Expert Witnesses Allowed to Use AI in Reports?\nGenerative AI is not inherently disqualifying for expert work, and there is a legitimate case for allowing measured use. Experts can use AI to play devil’s advocate, run mock cross-examinations or stress-test a report against an opposing expert’s likely positions, and used well, these tools can genuinely strengthen the rigor of an opinion.\nThe key is ensuring the expert can demonstrate that the final opinion reflects independent professional judgment rather than AI output. An expert should be prepared to show that AI-generated material was verified against the underlying facts, that the expert’s own expertise shaped the ultimate conclusions and that clear documentation connects the evidence to the opinion offered at trial.\nWhat Practical Steps Should Litigation Teams Take Now?\nCounsel retaining experts should assume that opposing counsel will request AI conversations and that those conversations, if left unmanaged, could seriously undermine the expert’s credibility at deposition or trial. Raising this issue with experts early and often, rather than waiting until AI use surfaces in discovery, is now a basic component of expert witness management.\nLitigation teams should build AI considerations into broader discovery strategy, both offensively and defensively, rather than treating AI use as an issue that only applies to their own experts. Requesting an opposing expert’s AI logs can be as valuable as protecting one’s own expert’s process, and teams that treat AI discovery as routine will be better positioned than those still treating it as a novel issue.\nHow Should Litigation Teams Build AI Into Their Discovery Strategy?\nGenerative AI has already moved from a theoretical discovery issue to a real source of trial exposure for expert witnesses. Courts are beginning to order disclosure of AI prompts and outputs, and existing Rule 29 stipulations covering expert notes and communications are proving insufficient to shield this new category of record.\nCompanies and counsel that update expert engagement letters now, with clear disclosure, approval, recordkeeping and verification requirements, will be far better positioned than those that wait until an AI chat log surfaces during a deposition. As generative AI becomes further embedded in expert practice, proactive governance of its use will increasingly separate credible expert testimony from testimony vulnerable to serious challenge.","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//natlawreview.com/article/ai-expert-witness-testimony-what-litigators-need-know-about-llm-discovery-risks","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 6987 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 6987 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":6987,"summary_length":375,"usable_text_length":6987,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":6987,"summary_length":375}},"tags":[],"format_contract_version":"news_item_formats.v1"}}}