# Hallucination indemnification in AI vendor contracts[^about]
How AI vendor contracts handle hallucination losses, output indemnity, IP claims, exclusions, court sanctions, and risk allocation.
## Do AI vendors indemnify customers for hallucinated output losses? {#do-ai-vendors-indemnify-hallucinations}
**Short answer.** Usually no. Public AI contracts mostly cover third-party IP claims, while ordinary losses from false or inaccurate output usually remain with the customer.
The market answer, as of April 2026, is fairly crisp. Public AI contracts increasingly indemnify customers against third-party copyright or broader IP claims tied to model output, but they do not generally indemnify customers for hallucinations as such. [^openai-service-terms-and-openai-services-agreeme][^google-cloud-service-specific-terms][^anthropic-on-bedrock-commercial-terms-of-service][^microsoft-customer-copyright-commitment-required] The same contracts usually say output may be inaccurate, require review, and put accuracy evaluation on the customer."solely responsible"[^openai-openai-services-agreement] for evaluating output; Google says models "may provide inaccurate or offensive Generated Output"[^google-cloud-service-specific-terms]; Anthropic warns outputs may be "false, incomplete, misleading"[^anthropic-on-bedrock-commercial-terms-of-service]. [^microsoft-turn-on-ai-disclaimers-in-microsoft-36] So the important split is not `AI indemnity` versus `no AI indemnity`. It is third-party IP claim versus everything else. If the bad output triggers an infringement claim, coverage may exist. If the bad output causes a bad filing, a false customer promise, a defamation theory, or a compliance error, public terms usually leave that risk with the deployer. [^wilson-sonsini-will-indemnification-commitments][^morgan-lewis-contract-corner-ensuring-ip-provisi][^cleary-gottlieb-managing-ai-risk-legal-and-gover]
## What cases show who bears losses from AI hallucinations? {#cases-allocate-ai-hallucination-loss}
**Short answer.** Unclear, but current public cases point to ordinary contract, tort, misrepresentation, defamation, and professional-duty rules. They do not create a default customer indemnity right against the AI vendor.
No statute or reported case in the source set creates a default rule that an AI vendor must indemnify a customer for hallucinated output. The governing law is still ordinary contract allocation, with tort, misrepresentation, defamation, and professional-responsibility doctrines in the background. [^mata-v-avianca-inc-678-f-supp-3d-443-s-d-n-y-202][^reuters-openai-defeats-radio-host-s-lawsuit-over][^mccarthy-t-trault-moffatt-v-air-canada-a-misrepr]
`Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)` remains the clearest judicial statement of the baseline. Judge Castel wrote that "existing rules impose a gatekeeping role on attorneys"[^mata-v-avianca-inc-678-f-supp-3d-443-s-d-n-y-202] to ensure accuracy. [^mata-v-avianca-inc-678-f-supp-3d-443-s-d-n-y-202] That was not an indemnity case. But it is a strong signal that courts can assign hallucination loss to the user when verification duties are ignored.
The next cases point in the same direction, though more indirectly. Public reporting on `Walters v. OpenAI, L.L.C.` says a Georgia court granted summary judgment to OpenAI in a defamation suit over false ChatGPT output, with the reporting emphasizing the role of accuracy warnings and the plaintiff's difficulty proving defamation on those facts. [^reuters-openai-defeats-radio-host-s-lawsuit-over][^quinn-emanuel-artificial-intelligence-update-apr] Public commentary on `Moffatt v. Air Canada` reads the British Columbia tribunal decision the other way: the company remained responsible for a chatbot's incorrect statement about bereavement fares. [^american-bar-association-business-law-today-bc-t][^mccarthy-t-trault-moffatt-v-air-canada-a-misrepr] Put together, those cases suggest a narrow but useful point. Courts have not used hallucination doctrine to create a new customer indemnity right. They are still allocating these disputes through older doctrines.
The more notable primary-law fact is the absence of precedent. The source set did not surface a published decision where a customer publicly invoked and litigated a hallucination-specific indemnity against a model provider. The field is being shaped in procurement and product terms, not yet in reported indemnity disputes.
## What do law firms say about AI output indemnity coverage? {#law-firms-view-ai-output-indemnity}
**Short answer.** Usually, law firm commentary treats public AI output indemnity as a narrow and heavily conditioned market response. The coverage is most developed for IP claims, not for business losses caused by hallucinated output.
The firms are closer together than the marketing language makes it seem. Wilson Sonsini says providers are offering indemnification because customers demand it, but also says many remain hesitant because of training-data claims, prompt-driven infringement, and the unpredictability of general-purpose models. [^wilson-sonsini-will-indemnification-commitments-2] Morgan Lewis says output indemnities are becoming more common, especially after hyperscalers moved first, but stresses that their scope can vary greatly and arrive with heavy carve-outs. [^morgan-lewis-contract-corner-ensuring-ip-provisi-2]
Orrick and Baker McKenzie are useful because they treat the problem as commercial fit rather than as a new body of law. Orrick says AI providers will likely resist the warranties and indemnities customers expect in ordinary software deals, while also treating hallucinations as a separate challenge rather than a covered risk. [^orrick-8-intellectual-property-and-commercial-qu] Baker McKenzie makes a similar point from the buyer side: ownership, risk, and indemnification all become harder to allocate when the output itself can be flawed or biased and standard SaaS clauses do not fit neatly. [^baker-mckenzie-innovation-and-accountability-ask]
Cleary Gottlieb and Fisher Phillips are the bluntest. Cleary says typical AI vendor contracts often limit or exclude indemnification for AI-generated output, leaving the organization rather than the vendor responsible. [^cleary-gottlieb-managing-ai-risk-legal-and-gover-2][^cleary-gottlieb-generative-ai-practical-consider] Fisher Phillips says some vendor contracts leave businesses using AI with the risk of the chatbot's conduct, even when the underlying technology was supplied by someone else. [^fisher-phillips-ai-hallucinations-could-cause-ni][^fisher-phillips-10-biggest-mistakes-businesses-m]
The interesting disagreement is not whether hallucination indemnity exists today. It is what sits in its place. Some firms focus on audit rights, testing records, and output controls. Others focus on human review and deployment limits. But none of the cited firms describes a public market where vendors routinely defend customers against losses caused by false output as such. [^wilson-sonsini-will-indemnification-commitments-2][^orrick-8-intellectual-property-and-commercial-qu][^morgan-lewis-contract-corner-ensuring-ip-provisi-2][^baker-mckenzie-innovation-and-accountability-ask][^cleary-gottlieb-managing-ai-risk-legal-and-gover-2][^fisher-phillips-ai-hallucinations-could-cause-ni]
## How do AI contract exclusions narrow output indemnity coverage? {#ai-contract-exclusions-narrow-output-indemnity}
**Short answer.** Usually through scope and conditions, not just liability caps. Public terms commonly narrow coverage by claim type, input and output handling, safety controls, service tier, and required mitigations.
- `Copyright shield` is narrower than it sounds. In public AI contracts it usually protects against third-party IP claims, not against first-party business loss from wrong output. [^openai-service-terms-and-openai-services-agreeme-2][^google-cloud-generative-ai-indemnified-services][^anthropic-on-bedrock-commercial-terms-of-service-2][^microsoft-customer-copyright-commitment-required-2]
- The real allocation device is exclusions. Public terms tend to withdraw coverage for modified output, customer-provided or unlicensed inputs, failure to use citations, filters, or guardrails, trademark use, third-party offerings, or preview and free tiers. [^openai-service-terms-and-openai-services-agreeme-2][^google-cloud-generative-ai-indemnified-services][^anthropic-on-bedrock-commercial-terms-of-service-2][^microsoft-customer-copyright-commitment-required-2]
- For the largest vendors, the narrowing mechanism is often not a tiny fee cap. OpenAI says its service-term indemnity is not subject to any liability cap; Google classifies indemnification as an `Unlimited Liability`; Anthropic says its fee cap does not apply to indemnification; Microsoft places defense obligations outside the ordinary cap structure. [^openai-openai-services-agreement-2][^google-cloud-terms-of-service][^anthropic-on-bedrock-commercial-terms-of-service-2][^microsoft-online-subscription-agreement] For top-tier public terms, the harder move is scoping and exclusion, not an obvious cap set at contract value.
## When can hallucinated AI output trigger an IP indemnity claim? {#hallucinated-ai-output-ip-indemnity}
**Short answer.** It depends on whether the false output also creates a third-party IP claim. A hallucination that causes a bad filing, customer promise, defamation allegation, or compliance error usually sits outside public IP indemnity language.
- Accuracy language points the other direction. Google says its services are not designed for or intended to meet Customer's regulatory, legal, or other obligations. Microsoft now offers AI disclaimers rather than hallucination indemnity. [^google-cloud-service-specific-terms-2][^microsoft-turn-on-ai-disclaimers-in-microsoft-36-2] The more the workflow touches legal, medical, regulatory, or customer-facing commitments, the more the contract paper reads as warning text rather than assurance. [^openai-service-terms-and-openai-services-agreeme-3][^fisher-phillips-10-biggest-mistakes-businesses-m-2]
- The boundary case is when a hallucination also becomes an IP problem. If a model fabricates an answer by reproducing copyrighted text, code, or imagery closely enough to trigger an infringement claim, the dispute may enter the IP-indemnity lane. If the same answer causes a bad filing, a false marketing claim, a customer-service promise, a defamation allegation, or a compliance error, public terms usually do not move the loss back to the vendor. [^openai-service-terms-and-openai-services-agreeme-3][^google-cloud-generative-ai-indemnified-services-2][^anthropic-on-bedrock-commercial-terms-of-service-3][^microsoft-customer-copyright-commitment-required-3]
- Perhaps the main unresolved question is whether courts will eventually treat systemic hallucination as a product-defect or architectural-negligence issue rather than as ordinary user-side verification failure. The current public cases still sit mostly inside negligence, defamation, misrepresentation, and professional-duty frameworks. [^mata-v-avianca-inc-678-f-supp-3d-443-s-d-n-y-202-2][^reuters-openai-defeats-radio-host-s-lawsuit-over-2][^mccarthy-t-trault-moffatt-v-air-canada-a-misrepr-2]
- It is still unsettled how much disclaimer language does. `Walters` suggests warnings may matter a great deal. `Moffatt` suggests a chatbot warning does not erase responsibility when the company itself is making a representation through the product. [^reuters-openai-defeats-radio-host-s-lawsuit-over-2][^american-bar-association-business-law-today-bc-t-2]
- The contract vocabulary may matter less than people think. Public terms rarely use `hallucination`; they speak in `inaccurate`, `false`, `incomplete`, or `misleading` output. That broader category could matter more than the label in future disputes. [^google-cloud-service-specific-terms-2][^anthropic-on-bedrock-commercial-terms-of-service-3][^microsoft-transparency-note-for-azure-openai]
- It is also unsettled whether broader output indemnity will appear as models become more agentic. The current movement is narrower: stronger IP shields, more conditions, more safety-tool requirements, and more explicit review obligations. [^wilson-sonsini-will-indemnification-commitments-3][^orrick-8-intellectual-property-and-commercial-qu-2][^morgan-lewis-contract-corner-ensuring-ip-provisi-3]
- And no one has yet shown what a litigated hallucination indemnity looks like. Perhaps those disputes are happening in private procurement. They have not yet produced a visible body of precedent.
[^about]: By Steven Obiajulu, J.D. Published by [openagreements.org](https://openagreements.org). Last reviewed 2026-04-20. License: CC BY 4.0. Steven Obiajulu, J.D. edits this topic article for Federal + multi-state coverage. It synthesizes legal sources and is not legal advice. This article is for informational purposes only and does not create an attorney-client relationship.
[^openai-service-terms-and-openai-services-agreeme]: **OpenAI, Service terms and OpenAI Services Agreement** — "OpenAI’s indemnification obligations to API customers under the Agreement include any third party claim that Customer’s use or distribution of Output infringes a third party’s intellectual property right." *OpenAI, Service terms and OpenAI Services Agreement.*
[^google-cloud-service-specific-terms]: **Google Cloud, Service Specific Terms** — "may provide inaccurate or offensive Generated Output" *Google Cloud, Service Specific Terms.*
[^anthropic-on-bedrock-commercial-terms-of-service]: **Anthropic on Bedrock, Commercial Terms of Service** — "false, incomplete, misleading" *Anthropic on Bedrock, Commercial Terms of Service.*
[^microsoft-customer-copyright-commitment-required]: **Microsoft, Customer Copyright Commitment Required Mitigations** — "The Customer Copyright Commitment (‘CCC’) is a provision in the Microsoft Product Terms that describes Microsoft's obligation to defend customers against certain third-party intellectual property claims relating to Output Content." *Microsoft, Customer Copyright Commitment Required Mitigations.*
[^openai-openai-services-agreement]: **OpenAI, OpenAI Services Agreement** — "solely responsible" *OpenAI, OpenAI Services Agreement.*
[^microsoft-turn-on-ai-disclaimers-in-microsoft-36]: **Microsoft, Turn on AI Disclaimers in Microsoft 365 Copilot** — "As the administrator of a Microsoft 365 organization, you can turn on Microsoft 365 Copilot AI disclaimer messages across supported applications." *Microsoft, Turn on AI Disclaimers in Microsoft 365 Copilot.*
[^wilson-sonsini-will-indemnification-commitments]: **Wilson Sonsini commentary** — "In response to the infringement fears, some generative AI providers are trying to get in front of the issue by offering to protect their customers from claims through indemnification." *Wilson Sonsini, Will Indemnification Commitments Address Market Demands in AI?.*
[^morgan-lewis-contract-corner-ensuring-ip-provisi]: **Morgan Lewis commentary** — "it is important that contracts relating to the use of GenAI and its outputs address the ownership/licensing of such GenAI outputs in order to document the agreement of the parties in the absence of legislative protections." *Morgan Lewis, Contract Corner: Ensuring IP Provisions Are Fit for GenAI.*
[^cleary-gottlieb-managing-ai-risk-legal-and-gover]: **Cleary Gottlieb commentary** — "Regulators have made clear that businesses deploying AI remain fully accountable for legal compliance, even where AI functionality is sourced from third-party vendors." *Cleary Gottlieb, Managing AI Risk: Legal and Governance Imperatives for the Board.*
[^mata-v-avianca-inc-678-f-supp-3d-443-s-d-n-y-202]: **Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)** — "existing rules impose a gatekeeping role on attorneys" *Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023).*
[^reuters-openai-defeats-radio-host-s-lawsuit-over]: **Reuters, OpenAI defeats radio host's lawsuit over allegations invented by ChatGPT** — "Judge Tracie Cason of Gwinnett County Superior Court ruled that plaintiff Mark Walters had not shown he was defamed and said OpenAI's chatbot ChatGPT puts users on notice that it can make errors." *Reuters, OpenAI defeats radio host's lawsuit over allegations invented by ChatGPT.*
[^mccarthy-t-trault-moffatt-v-air-canada-a-misrepr]: **McCarthy Tétrault, Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot** — "The decision held that a company can be liable for negligent misrepresentations made by a chatbot on a publicly available commercial website." *McCarthy Tétrault, Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot.*
[^quinn-emanuel-artificial-intelligence-update-apr]: **Quinn Emanuel, Artificial Intelligence Update - April 2026** — "the first wave of cases suggests that many disputes involving conversational AI are being litigated on theories that do not depend on Section 230 at all." *Quinn Emanuel, Artificial Intelligence Update - April 2026.*
[^american-bar-association-business-law-today-bc-t]: **American Bar Association Business Law Today, BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot** — "the Tribunal found that Air Canada still bore responsibility for all the information on its website, whether it came from a static page or a chatbot." *American Bar Association Business Law Today, BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot.*
[^wilson-sonsini-will-indemnification-commitments-2]: **Wilson Sonsini commentary** — "In response to the infringement fears, some generative AI providers are trying to get in front of the issue by offering to protect their customers from claims through indemnification." *Wilson Sonsini, Will Indemnification Commitments Address Market Demands in AI?.*
[^morgan-lewis-contract-corner-ensuring-ip-provisi-2]: **Morgan Lewis commentary** — "it is important that contracts relating to the use of GenAI and its outputs address the ownership/licensing of such GenAI outputs in order to document the agreement of the parties in the absence of legislative protections." *Morgan Lewis, Contract Corner: Ensuring IP Provisions Are Fit for GenAI.*
[^orrick-8-intellectual-property-and-commercial-qu]: **Orrick commentary** — "Companies are not immune from legal, regulatory and ethical concerns created by generative AI tools merely because they are a ‘user’ and not the ‘developer.’" *Orrick, 8 Intellectual Property and Commercial Questions to Ask Your Generative AI Tool Provider.*
[^baker-mckenzie-innovation-and-accountability-ask]: **Baker McKenzie commentary** — "A company that does not have proper safeguards around GAI and does not ensure that contractors or vendors with access to its data use safeguards, arguably is not making reasonable efforts to protect its information and reduce the risk of trade secret disclosure." *Baker McKenzie, Innovation and Accountability: Asking Better Questions in Implementing Generative AI.*
[^cleary-gottlieb-managing-ai-risk-legal-and-gover-2]: **Cleary Gottlieb commentary** — "Regulators have made clear that businesses deploying AI remain fully accountable for legal compliance, even where AI functionality is sourced from third-party vendors." *Cleary Gottlieb, Managing AI Risk: Legal and Governance Imperatives for the Board.*
[^cleary-gottlieb-generative-ai-practical-consider]: **Cleary Gottlieb, Generative AI: Practical Considerations for Companies and Boards** — "At any depth of deployment, companies and boards need to be aware of key risks AI poses and areas of uncertainty in the laws governing its use." *Cleary Gottlieb, Generative AI: Practical Considerations for Companies and Boards.*
[^fisher-phillips-ai-hallucinations-could-cause-ni]: **Fisher Phillips commentary** — "AI hallucinations are confidently incorrect outputs generated by large language models (LLMs). They happen because GenAI is designed to predict the most likely next word, not to verify facts." *Fisher Phillips, AI Hallucinations Could Cause Nightmares for Your Business: 10 Steps You Can Take to Safeguard Your GenAI Use.*
[^fisher-phillips-10-biggest-mistakes-businesses-m]: **Fisher Phillips, 10 Biggest Mistakes Businesses Make When Deploying AI Chatbots - And 10 Fixes You Can Make Today** — "Under every chatbot law enacted to date, the deploying business carries legal responsibility for what the chatbot does, regardless of who developed it." *Fisher Phillips, 10 Biggest Mistakes Businesses Make When Deploying AI Chatbots - And 10 Fixes You Can Make Today.*
[^openai-service-terms-and-openai-services-agreeme-2]: **OpenAI, Service terms and OpenAI Services Agreement** — "OpenAI’s indemnification obligations to API customers under the Agreement include any third party claim that Customer’s use or distribution of Output infringes a third party’s intellectual property right." *OpenAI, Service terms and OpenAI Services Agreement.*
[^google-cloud-generative-ai-indemnified-services]: **Google Cloud, Generative AI Indemnified Services** — "The following Services are covered by the Additional Google Indemnification Obligations in the Service Specific Terms." *Google Cloud, Generative AI Indemnified Services.*
[^anthropic-on-bedrock-commercial-terms-of-service-2]: **Anthropic on Bedrock, Commercial Terms of Service** — "false, incomplete, misleading" *Anthropic on Bedrock, Commercial Terms of Service.*
[^microsoft-customer-copyright-commitment-required-2]: **Microsoft, Customer Copyright Commitment Required Mitigations** — "The Customer Copyright Commitment (‘CCC’) is a provision in the Microsoft Product Terms that describes Microsoft's obligation to defend customers against certain third-party intellectual property claims relating to Output Content." *Microsoft, Customer Copyright Commitment Required Mitigations.*
[^openai-openai-services-agreement-2]: **OpenAI, OpenAI Services Agreement** — "solely responsible" *OpenAI, OpenAI Services Agreement.*
[^google-cloud-terms-of-service]: **Google Cloud Terms of Service** — "These Google Cloud Terms of Service (together, the ‘Agreement’) are entered into by Google and the entity or person agreeing to these terms (‘Customer’) and govern Customer's access to and use of the Services." *Google Cloud Terms of Service.*
[^microsoft-online-subscription-agreement]: **Microsoft, Online Subscription Agreement** — "The aggregate liability of each party for all claims under this agreement is limited to direct damages up to the amount paid under this agreement for the Online Service during the 12 months before the cause of action arose" *Microsoft, Online Subscription Agreement.*
[^google-cloud-service-specific-terms-2]: **Google Cloud, Service Specific Terms** — "may provide inaccurate or offensive Generated Output" *Google Cloud, Service Specific Terms.*
[^microsoft-turn-on-ai-disclaimers-in-microsoft-36-2]: **Microsoft, Turn on AI Disclaimers in Microsoft 365 Copilot** — "As the administrator of a Microsoft 365 organization, you can turn on Microsoft 365 Copilot AI disclaimer messages across supported applications." *Microsoft, Turn on AI Disclaimers in Microsoft 365 Copilot.*
[^openai-service-terms-and-openai-services-agreeme-3]: **OpenAI, Service terms and OpenAI Services Agreement** — "OpenAI’s indemnification obligations to API customers under the Agreement include any third party claim that Customer’s use or distribution of Output infringes a third party’s intellectual property right." *OpenAI, Service terms and OpenAI Services Agreement.*
[^fisher-phillips-10-biggest-mistakes-businesses-m-2]: **Fisher Phillips, 10 Biggest Mistakes Businesses Make When Deploying AI Chatbots - And 10 Fixes You Can Make Today** — "Under every chatbot law enacted to date, the deploying business carries legal responsibility for what the chatbot does, regardless of who developed it." *Fisher Phillips, 10 Biggest Mistakes Businesses Make When Deploying AI Chatbots - And 10 Fixes You Can Make Today.*
[^google-cloud-generative-ai-indemnified-services-2]: **Google Cloud, Generative AI Indemnified Services** — "The following Services are covered by the Additional Google Indemnification Obligations in the Service Specific Terms." *Google Cloud, Generative AI Indemnified Services.*
[^anthropic-on-bedrock-commercial-terms-of-service-3]: **Anthropic on Bedrock, Commercial Terms of Service** — "false, incomplete, misleading" *Anthropic on Bedrock, Commercial Terms of Service.*
[^microsoft-customer-copyright-commitment-required-3]: **Microsoft, Customer Copyright Commitment Required Mitigations** — "The Customer Copyright Commitment (‘CCC’) is a provision in the Microsoft Product Terms that describes Microsoft's obligation to defend customers against certain third-party intellectual property claims relating to Output Content." *Microsoft, Customer Copyright Commitment Required Mitigations.*
[^mata-v-avianca-inc-678-f-supp-3d-443-s-d-n-y-202-2]: **Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)** — "existing rules impose a gatekeeping role on attorneys" *Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023).*
[^reuters-openai-defeats-radio-host-s-lawsuit-over-2]: **Reuters, OpenAI defeats radio host's lawsuit over allegations invented by ChatGPT** — "Judge Tracie Cason of Gwinnett County Superior Court ruled that plaintiff Mark Walters had not shown he was defamed and said OpenAI's chatbot ChatGPT puts users on notice that it can make errors." *Reuters, OpenAI defeats radio host's lawsuit over allegations invented by ChatGPT.*
[^mccarthy-t-trault-moffatt-v-air-canada-a-misrepr-2]: **McCarthy Tétrault, Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot** — "The decision held that a company can be liable for negligent misrepresentations made by a chatbot on a publicly available commercial website." *McCarthy Tétrault, Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot.*
[^american-bar-association-business-law-today-bc-t-2]: **American Bar Association Business Law Today, BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot** — "the Tribunal found that Air Canada still bore responsibility for all the information on its website, whether it came from a static page or a chatbot." *American Bar Association Business Law Today, BC Tribunal Confirms Companies Remain Liable for Information Provided by AI Chatbot.*
[^microsoft-transparency-note-for-azure-openai]: **Microsoft, Transparency Note for Azure OpenAI** — "The service doesn't have information about events that occur after its training date, likely has missing knowledge about some topics, and may not always produce factually accurate information." *Microsoft, Transparency Note for Azure OpenAI.*
[^wilson-sonsini-will-indemnification-commitments-3]: **Wilson Sonsini commentary** — "In response to the infringement fears, some generative AI providers are trying to get in front of the issue by offering to protect their customers from claims through indemnification." *Wilson Sonsini, Will Indemnification Commitments Address Market Demands in AI?.*
[^orrick-8-intellectual-property-and-commercial-qu-2]: **Orrick commentary** — "Companies are not immune from legal, regulatory and ethical concerns created by generative AI tools merely because they are a ‘user’ and not the ‘developer.’" *Orrick, 8 Intellectual Property and Commercial Questions to Ask Your Generative AI Tool Provider.*
[^morgan-lewis-contract-corner-ensuring-ip-provisi-3]: **Morgan Lewis commentary** — "it is important that contracts relating to the use of GenAI and its outputs address the ownership/licensing of such GenAI outputs in order to document the agreement of the parties in the absence of legislative protections." *Morgan Lewis, Contract Corner: Ensuring IP Provisions Are Fit for GenAI.*