The NIST Generative Artificial Intelligence Profile: a useful issue-spotter for contracts

This week’s musings on tech contracts

Most of us know that generative artificial intelligence creates some unusual risks. Few of us, however, have taken the time to think through and articulate a complete list. Fortunately, the National Institute of Standards and Technology (NIST) did it it for us. Their twelve-point list of gen-AI risks offers an issue-spotter for contract-drafters, representing both customers and providers.

NIST’s 12-Point List as an Issue-Spotter

NIST published its Generative Artificial Intelligence Profile in July, as part of its larger Artificial Intelligence Risk Management Framework. (It’s all in response to President Biden’s October 2023 Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence.) The profile looks to me like an expert analysis of the risks, including a few you might not have thought of or at least thought through. So it offers us an issue spotter.

I love issue-spotters. It’s hard to know the answer to every contract or legal question, but if we spot the issue, we can look up the answer, usually. If we don’t spot the issue … well, it’s the “unknown unknown” as Donald Rumsfeld once put it, and we can’t address it.

Using the List: Gen-AI Provider and Customer

I suggest you check through the NIST list, summarized below, when you work on a gen-AI contract. Ask which risks arise for your side and what the contract could do about them. Here are some ideas and examples.

  • If you’re the gen-AI provider, your customer could fail to take precautions, leading to third party injuries. For instance, your customer could ignore risk #6, “harmful bias or homogenization,” and rely on outputs that get you sued by victims of discrimination. So you should consider contract terms prohibiting discrimination and other misuse of outputs, or at least requiring precautions, as well as indemnities against resulting lawsuits. You should also consider disclosures about the risk – e.g., “Customer is on notice that outputs may include harmful bias or Homogenization (as defined __), and Customer recognizes and agrees that it and not Provider is responsible for preventing of any resulting harm to third parties.”
  • If you’re the gen-AI customer, you could be buying a system with high odds of mishandling data or generating misleading outputs. For instance, the system could suffer more than your realize from risk #2, “confabulation” – usually called “hallucination” (as if a computer could imagine and dream). So you should consider SLAs and other terms (maybe warranties) promising that the system won’t confabulate – or won’t confabulate more than X% of the time. You should also consider indemnities against third party suits resulting from confabulation, as well as disclosures or even representations about confabulation frequency. – e.g., “Vendor represents that, to the best of its knowledge, the System’s Confabulation Frequency (as defined __) is no more than Y%.”

I’m not saying you’ll get all these terms from the other side. Your odds depend on the system, including the nature of its training data. But you have no chance if you don’t spot the issue.

Your Issue-Spotter

None of us likes too much reading, so I’ve pasted NIST’s twelve-risk summary below.

  1. CBRN Information or Capabilities: Eased access to or synthesis of materially nefarious information or design capabilities related to chemical, biological, radiological, or nuclear (CBRN) weapons or other dangerous materials or agents.
  2. Confabulation: The production of confidently stated but erroneous or false content (known colloquially as “hallucinations” or “fabrications”) by which users may be misled or deceived.
  3. Dangerous, Violent, or Hateful Content: Eased production of and access to violent, inciting, radicalizing, or threatening content as well as recommendations to carry out self-harm or conduct illegal activities. Includes difficulty controlling public exposure to hateful and disparaging or stereotyping content.
  4. Data Privacy: Impacts due to leakage and unauthorized use, disclosure, or de-anonymization of biometric, health, location, or other personally identifiable information or sensitive data.
  5. Environmental Impacts: Impacts due to high compute resource utilization in training or operating GAI models, and related outcomes that may adversely impact ecosystems.
  6. Harmful Bias or Homogenization: Amplification and exacerbation of historical, societal, and systemic biases; performance disparities between sub-groups or languages, possibly due to non-representative training data, that result in discrimination, amplification of biases, or incorrect presumptions about performance; undesired homogeneity that skews system or model outputs, which may be erroneous, lead to ill-founded decision-making, or amplify harmful biases.
  7. Human-AI Configuration: Arrangements of or interactions between a human and an AI system which can result in the human inappropriately anthropomorphizing GAI systems or experiencing algorithmic aversion, automation bias, over-reliance, or emotional entanglement with GAI systems.
  8. Information Integrity: Lowered barrier to entry to generate and support the exchange and consumption of content which may not distinguish fact from opinion or fiction or acknowledge uncertainties, or could be leveraged for large-scale dis- and mis-information campaigns.
  9. Information Security: Lowered barriers for offensive cyber capabilities, including via automated discovery and exploitation of vulnerabilities to ease hacking, malware, phishing, offensive cyber operations, or other cyberattacks; increased attack surface for targeted cyberattacks, which may compromise a system’s availability or the confidentiality or integrity of training data, code, or model weights.
  10. Intellectual Property: Eased production or replication of alleged copyrighted, trademarked, or licensed content without authorization (possibly in situations which do not fall under fair use); eased exposure of trade secrets; or plagiarism or illegal replication.
  11. Obscene, Degrading, and/or Abusive Content: Eased production of and access to obscene, degrading, and/or abusive imagery which can cause harm, including synthetic child sexual abuse material (CSAM), and nonconsensual intimate images (NCII) of adults.
  12. Value Chain and Component Integration: Non-transparent or untraceable integration of upstream third-party components, including data that has been improperly obtained or not processed and cleaned due to increased automation from GAI; improper supplier vetting across the AI lifecycle; or other issues that diminish transparency or accountability for downstream users.

If you’d like to learn more about AI contracts, check out our course, Artificial Intelligence Contracts: Drafting and Negotiating. And check out The Tech Contracts Master Class™, which covers all the key terms in IT agreements, including contracts for AI.


© 2024 by Tech Contracts Academy, LLC. All rights reserved. Thank you to Pixabay.com for great, free stock photos!

Related Posts