Hyperlinked Documents in Discovery: Are they Attachments?
What is a hyperlinked file and is it an “attachment?”
- It has become a widely accepted principle in e-discovery that a static document or file attached to an email creates a document “family” where the email is the “parent” and the attachment is the “child.” The relationship between the attachment and the email is easily understood to be one unit that should be preserved in document production.
- New communication platforms have changed the way documents can be shared. Traditionally, we send a copy of a document via email which represented a static snapshot of that document as it existed at the time the email was sent. Now, documents can be shared and edited in “real-time”, so that we don’t have to save a new copy of a document every time someone edits it and it is accessible to recipients in the revised form in real time. We share an active “link” to that document instead. This is a hyperlinked document or file.
- Should a hyperlinked file be treated the same as an attachment to a parent email in discovery? How does this impact preservation efforts? If the email containing the link becomes discoverable, are we obligated to collect the version of the document associated with the link as it was at the time that email was sent? As it exists at the time of collection? Sometime in between? Or all?
- The process of collecting a file referenced in a hyperlink cannot yet be automated like collecting a copy of a file attached to an email.
Do the Federal Rules of Civil Procedure address this issue?
- Yes and no. The rules and the Rules Committee commentary do not address hyperlinked documents by name, but are built on relevant principles. The Committee commentary points practitioners to Rules 26 and 34. Courts are increasingly being asked to rule on how hyperlinks should be treated by producing parties when disputes arise between parties based upon the principles.
-
Rule 26(b)(1): The scope of discovery extends to any “nonprivileged matter” that is “relevant” and “proportional to the needs of the case.” Courts have recently used a “proportionality” standard when evaluating the duties and burdens of producing parties with respect to hyperlinked files.
-
Rule 34: Rule 34 rule governs the scope and procedure by which a party in a legal proceeding can request access to documents, ESI, and tangible items held by another party. The rule also requires that documents be produced “as they are kept in the usual course of business.” According to interpretations of the rule ESI attachments are generally required to be produced along with the relevant parent email as a family unit to reflect how the documents were stored in the “usual course of business.” However, the rule does not define an “attachment”.
Recent rulings addressing hyperlinks and ESI
| Case |
Core Issue |
Ruling |
Key Takeaways |
|
Nicholas v. Noom 2021 LEXIS 46860, 2021 WL 94860 (S.D.N.Y. March 11, 2021) (reconsidering prior order)
|
Is a hyperlinked file the same as an attachment? |
A hyperlinked file is not the same as an attachment under Rule 34 based on the “proportionality” standard in Rule 26(b)(1). The burden of collecting hyperlinks was extremely high. Plaintiffs failed to demonstrate adequate need or relevance. |
ESI Protocols should define “attachments” and parties must show relevance of hyperlinked data. (The parties' ESI protocol entered as an order did not define attachment.) |
|
In re Acetaminophen, No. 22md3043 (DLC), 2023 LX 86437 (S.D.N.Y. Jan. 17, 2023)
|
Are there situations where hyperlinked content might be treated like traditional “attachments?” |
Yes, the Court ruled hyperlinked files should be treated like traditional “attachments.” |
Courts expect litigants to be more precise in defining scope and burden (proportionality – explain why burdensome). |
|
McLaughlin v. Tesla, Inc., No. 22-cv-07849-SVK, 2023 U.S. Dist. LEXIS 233711 (N.D. Cal. Nov. 2, 2023)
In re Meta Pixel Healthcare Litig., No. 22-cv-03580-WHO (VKD), 2023 LX 57485 (N.D. Cal. June 2, 2023)
|
Are courts treating requests for production of hyperlinks with an “all or nothing” approach? |
No, in Tesla, the Court required production of 25 hyperlinks and rejected a blanket exclusion in favor of “measured and practical compliance.”
In Meta Pixel, hyperlinks were not treated as attachments by default. The Court allowed specific requests based on relevance and burden.
|
In 2023, courts began crafting middle-ground rulings balancing production burdens with discovery obligations. Parties should consider agreeing to produce a reasonable number of hyperlinked attachments. |
| In re StubHub, 2024 U.S. Dist. LEXIS 90202, 2024 WL 2305604 (N.D. Cal. May 20, 2024) |
The parties agreed to an ESI Order defining “families” as both traditional attachments and hyperlinked data--an agreement the court described as done "carelessly" by the defendant who'd been "foolish" to agreed to it without adequately investigating its feasability. During discovery, defendants ralized they couldn’t comply and moved to amend the protocol. They spent hundreds of hours and money hiring an outside vendor to identify correct versions of hyperlinked attachments and link them to their parent email. Plaintiff argued it wasn’t sufficient because many emails with “links” were still not associated with a produced attachment and sought sanctions. |
The ESI Protocol agreed to by both parties failed due to infeasibility (even though it included well-defined definitions of attachments, the definitions weren’t practical to comply with). The Court allowed defendants to modify the ESI protocol. The court denied sanctions, principally because the defendants' experts persuaded the Court compliance of the ESI protocol ordered was never possible. The Court emphasized need for due diligence before agreement. |
Potential Sanctions: Cautionary tale about overbroad and extensive ESI protocols without confirming the burde of complying and indeed, whether compliance is possible. Parties should anticipate that courts will hold them to their agreement. |
| In Re Uber Technologies (N.D. Cal. April 2024) |
A dispute arose about a provision in the ESI Protocol requiring the production of hyperlinks as attachments when plaintiffs determined that defendants’ production of hyperlinked documents were not the same versions linked in the messages. Defendant used Google Vault, which the defendant learned “does not automatically capture the real-time version of a hyperlinked document.” It only captures the current version at the time it is exported. Defendant argued that no proposed solution to try and marry up the version of hyperlinked files sent at the time of the message was possible without significant burden. |
The Court didn’t order production of all hyperlinks as agreed to. However, it noted it did not fully accept the defendant’s burden arguments because the defendant willingly chose to use Google Vault for storage and export despite its widely known technical limitations. |
Know the limitations of your client’s platform used for storage, collection, and export of ESI before agreeing to an ESI Protocol. |
PRACTICAL RECOMMENDATIONS FOR ESI PROTOCOLS
Define “Attachment”
- Avoid ambiguity (what does it include/exclude?)
- Tech Consideration: Clarify if the definition of “attachment” extends to modern attachments (link to live docs) or just static files
Differentiate Hyperlink Types
- Internal v. external (public) links
Dynamic Content and Versioning (Documenting burdens, limitations and processes)
- Ask clients if it is customary to use hyperlinks when communicating internally via email or a chat platform such as Teams.
- If so, ask clients to work with their IT department to determine: 1) if it is possible to preserve hyperlink version histories and capture dynamic content changes to hyperlinked files, and 2) what is the burden to do so? We should not be making blanket recommendations to clients that they must preserve and produce this data in all circumstances solely because it is possible to do so. Each client/matter will require analyzing and weighing the accessibility and relevance of requested hyperlinked data with the burden of preserving and collecting it. Once we and the client have an understanding of their hyperlink usage and preservation capabilities, we should advise the client on how best to reference and preserve hyperlinked data in legal hold or data
retention policies. The MVA e-Discovery team can assist you and our clients with drafting legal hold or retention policies to accurately document the processes and burdens that does not obligate clients to preserve all hyperlinked data in all circumstances or otherwise disadvantage clients in current or future discovery.
- Talk to tech folks: If the client does not have existing documentation regarding the processes to preserve and collect hyperlinked data, ask client IT contacts to prepare sample language to include from the specific platform they use for document management and collection as this process varies from platform to platform. Thus, you will not be able to reuse this language in an ESI protocol for different clients using different platforms.
Proportionality Assessment
- Weigh cost/burden against relevance to case: Remember each client/matter will require a separate analysis. Even if a client’s software has the capability of preserving multiple versions of hyperlinks, it is important to analyze and weigh the burdens of collection and production with the relevance to the matter so that both parties can reach agreement on what is mutually reasonable and within the proper scope of relevant information to produce under the applicable Rules and caselaw in the jurisdiction.
- Case Law: Proportionality concerns in Noom
Feasibility & Testing
- Conduct Feasibility testing before stipulating to hyperlink production
-
StubHub's unfeasible protocol demonstrates the risks of not doing so.
Prepare for the Rule 26(f) conference
- You need to investigate and fully understand how your client uses hyperlinks and how they can be preserved for production and be prepared to ask the other side the same questions far in advance of the Rule 26(f)conference.
- Courts want you to talk about this upfront, and if you don’t, they will have little sympathy if you didn’t do your homework before the 26(f) conference.
HANDLING HYPERLINKS IN DOCUMENT REQUESTS
Producing Party: How do you respond to a document request for “all hyperlinked files?”
- Before agreeing to any production or protocol, speak with the client to understand and estimate the likely volume of hyperlinks in relevant data. Understanding if “most” emails have hyperlinks or that hyperlinks are “rarely” used in emails is better than having no knowledge of how frequently they are used. The requesting party will likely request “all versions you have” if asked. Armed with an understanding of the likely volume of hyperlinks and the client’s burdens to preserve them, you can formulate a reasonable approach for discussion with requesting counsel about the scope of their request, including reciprocity for their hyperlinks.
Requesting Party: Considerations before drafting a request for hyperlinked data.
- Understand the scope of what you are requesting from the other party before drafting your request.
- Have a conversation with your client to learn its understanding of the opposing party’s hyperlink usage.
- Tailor your request to better describe its scope related to hyperlinks. For example, do you really want all (contemporaneous, current and updated) versions of a hyperlinked file? Will the court find it reasonable and proportional? Be careful of the reciprocal demand risk; can your client comply?
SAMPLE ESI PROTOCOL LANGUAGE FOR HYPERLINKED FILES
After receiving a production of documents responsive to properly propounded requests issued pursuant to Fed. R. Civ. P. 34 or 45, the Receiving Party may make reasonable and proportionate requests that the Producing Party conduct a reasonable search for internal or non-public documents identified via hyperlinks.
Collection and requests:
Hyperlinked Documents
Except as required in this paragraph, no Party shall be required to manually locate, collect, or associate a hyperlinked document with the message[1] containing the hyperlink. A Receiving Party[2] may request a manual search for a hyperlinked document that is likely to be material to the Party’s preparation of its case. Any such reasonable and particularized request shall provide a list of specifically requested hyperlinks (with the Beginning Bates and Bates pages on which the hyperlinks appear) by email to the Producing Party’s counsel, capped at 15 hyperlink inquiries[3] being open at any one time per Producing Party and no more than 100 total inquiries[4] per Producing Party across the duration of the litigation.
Qualifications and Clarifications:
The production of hyperlinked document shall not constitute an admission by the Producing Party that the produced document is substantively identical to the version of the document that existed at the time the message with the hyperlink was sent or received or that the produced document was viewed by the sender or recipient of the message. For the avoidance of doubt, hyperlinked documents shall not be construed to be in the same family as the message containing the hyperlink solely by virtue of the hyperlink reference.
To the extent any Producing Party has or can reasonably use electronic tools (such as Microsoft Purview eDiscovery Premium) to automatically collect the current, available version of non-public hyperlinked documents stored within the Microsoft 365 environment (or other comparable tools for their email platform), the Parties are willing to meet and confer as the process for potential searching, collection, and/or review using automated rather manual research.
[1] For clarity, the terms “message” or “messages” refer to email or Teams messages.
[2] For purposes of this provision, all MDL plaintiffs shall be treated as a single Receiving Party.
[3] Inquiries pursuant to this provision will be limited to hyperlinks referencing documents within the Producing Party’s Microsoft O365 or M365 environment.
[4] In some cases, the Producing Party may be unable to produce a hyperlinked document in response to a Receiving Party’s inquiry under this paragraph. Fifty (50) of such unsuccessful searches for a hyperlinked document shall not count against the 100-inquiry limitation on total inquiries per Producing Party described herein.
AI Prompts and Preservation
-
The New York Times Company, et al. v. Microsoft Corporation, et al, (S.D.N.Y. Case No. 1:23-CV-11195). In a landmark development, Magistrate Judge Ona T. Wang issued a sweeping preservation order in In re OpenAI, Inc. Copyright Infringement Litigation MDL 2025 U.S. Dist. LEXIS 97943, 2025 WL 1442678 (SDNY) requiring OpenAI to retain all ChatGPT output log data that would otherwise be deleted. This included user prompts and responses (even those users had deleted) raising profound implications for privacy, data governance, and eDiscovery.
- The preservation order issued on May 13, 2025 impacted over 400 million users worldwide and potentially 60+ billion chats, signaling a shift in AI data management. On June 26, 2025, Judge Sidney H. Stein denied OpenAI’s objection and left the order standing. AI generated content is now fair game in discovery and courts may lean towards broad preservation, especially in litigation alleging systemic misuse or pattern-based harm.
Defense Strategy: Preserving Smart, Pushing Back Hard
- Defense teams should act swiftly when litigation is reasonably anticipated, and AI tools are involved. Focus on:
-
Custodians: Identify employees using AI tools for work-related tasks
-
Platforms: Flag enterprise vs. consumer-grade AI usage (e.g., Copilot, ChatGPT, Google Gemini)
-
Content Types: Preserve prompts, responses, timestamps, and user metadata
-
Retention Settings: Use enterprise controls to retain or export logs
- Once the duty to preserve is triggered and you determine that AI prompts could be within scope, you should observe preservation best practices for AI tools such as:
-
Legal Hold Notices: Extend to relevant custodians using AI tools
-
Centralized Logging: Encourage enterprise AI use with built-in audit trails
-
Metadata Capture: Preserve context—who entered what, when, and why
-
Segregated Storage: Keep preserved AI data separate under litigation hold protocols
When & How to Push Back
- Preservation demands can be excessive. Defense teams should document their challenge of overreach due to reasons such as:
-
Overbreadth: Data volume is disproportionate to claims
-
Relevance: Most AI prompts are unrelated to litigation issues
-
Privacy: Preserving deleted chats may violate user expectations or data laws
-
Technical Burden: Ephemeral data (temporary information that exists only for a short period or for a specific purpose) may not be retrievable or only retrievable with undue cost
-
Enterprise Protections: AI use may be contractually shielded or excluded
- Use Rule 26(b)(1) (proportionality) and Rule 26(c) (protective orders) to limit the scope of preservation demands or requests for production. In the case of a fight over AI production, consider proposing (all on a reciprocal basis):
- Sampling protocols (sampling methodologies to validate the efficacy of preserving specific types of data)
- Temporal (or date-range) limitations
- Custodian-based filtering
- Exclusion of deleted or ephemeral data
Defense Playbook: AI in Discovery
-
Audit AI Usage: Know where and how your client uses generative AI and advise on litigation risks
-
Control the Narrative: Frame AI prompts as exploratory, not authoritative
-
Limit Exposure: Push for narrow relevance definitions when the definition may implicate AI as used by your client
-
Protect Privacy: Codify where appropriate and assert user expectations and regulatory constraints to object as well as foreign data protection standards that may apply or U.S. protections of personal identification information (PII)
-
Document Burden: Quantify cost and complexity of preservation and production
-
ESI Protocols: Set preservation limitations regarding generative AI
- For example: “Nothing in this Order shall require a party to preserve ephemeral AI data (e.g., prompts or outputs not retained by default) unless such data is reasonably accessible and relevant to the claims or defenses in this action. Parties may object to preservation demands on grounds of proportionality, undue burden, or privacy.”
- Also address third-party platform protections, if applicable. For example, “If any AI-Generated Content was created using third-party platforms, the producing party shall disclose whether such content is retrievable and whether any privacy or contractual limitations apply to its preservation or production. Parties may seek relief from the Court if such limitations materially affect discovery obligations.”
Looking Forward
- The NYT v. OpenAI decision signals a shift in judicial expectations regarding AI generated content. Defense teams must proactively assess AI usage, implement defensible preservation protocols, and be prepared to assert proportionality and burden-based objections when appropriate.
- For further assistance in drafting protective orders, ESI protocols, or preservation protocols, please contact the Discovery team.
- Discovery requests and interrogatories targeted to collect AI prompts and results (always consider the same requests coming back to you as Producing Party).
- “Artificial Intelligence Platform” means any software, model, or system incorporating artificial intelligence, including machine learning, natural language processing, large language model, or predictive analytics, used by you in relation to the Property, Complaint, or any allegation, claim or defense presented in this matter. Artificial Intelligence Platform includes but is not limited to programs such as Chat GPT, Grok, Co-Pilot, Gemini, Claude, or DeepSeek.
- Identify all prompts, questions, or other information inputted into any Artificial Intelligence Platform concerning the Property or any alleged issues you contend occurred or are occurring on the Property.
- Identify all prompts, questions, or other information inputted into any Artificial Intelligence Platform concerning the Complaint, this dispute, or any claims or defenses raised in this litigation.
- Produce your entire history for all Artificial Intelligence Platforms you have used for information, questions, or have otherwise consulted or used as justification for any demands or allegations concerning the Property, including but not limited to producing all inputs, prompts, responses, follow-up questions, and responses from the applicable platform.
- Produce your entire history for all Artificial Intelligence Platforms you have used for information, questions, or have otherwise consulted or used as justification for any demands or allegations concerning the Complaint, this dispute, or any claims or defenses raised in this litigation.
The Discovery Team is happy to review ESI Protocols and/or confidentiality agreements to ensure alignment with best practices. Please email any questions or requests for assistance to discovery@mvalaw.com.
|
