<?xml version="1.0" encoding="UTF-8"?>
<record
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://www.loc.gov/MARC21/slim http://www.loc.gov/standards/marcxml/schema/MARC21slim.xsd"
    xmlns="http://www.loc.gov/MARC21/slim">

  <leader>04007naa a22005297i 4500</leader>
  <controlfield tag="001">UMUN-OA-20261005-ACC-064</controlfield>
  <controlfield tag="003">local</controlfield>
  <controlfield tag="005">20261005174841.0</controlfield>
  <controlfield tag="006">m     o  d        </controlfield>
  <controlfield tag="007">cr||||||||||||</controlfield>
  <controlfield tag="008">261005s2023    xx      o    u000 0 eng d</controlfield>
  <datafield tag="024" ind1="7" ind2=" ">
    <subfield code="a">10.1007/s43681-023-00289-2</subfield>
    <subfield code="2">doi</subfield>
  </datafield>
  <datafield tag="035" ind1=" " ind2=" ">
    <subfield code="a">(local)OpenAlex:W4381572755</subfield>
  </datafield>
  <datafield tag="040" ind1=" " ind2=" ">
    <subfield code="a">local</subfield>
    <subfield code="b">eng</subfield>
    <subfield code="c">local</subfield>
  </datafield>
  <datafield tag="041" ind1="0" ind2=" ">
    <subfield code="a">eng</subfield>
  </datafield>
  <datafield tag="100" ind1="0" ind2=" ">
    <subfield code="a">Jakob M&#xF6;kander</subfield>
    <subfield code="e">author.</subfield>
  </datafield>
  <datafield tag="245" ind1="1" ind2="0">
    <subfield code="a">Auditing large language models: a three-layered approach /</subfield>
    <subfield code="c">Jakob M&#xF6;kander; Jonas Schuett; Hannah Rose Kirk [and others].</subfield>
  </datafield>
  <datafield tag="264" ind1=" " ind2="1">
    <subfield code="a">[Place of publication not identified] :</subfield>
    <subfield code="b">Springer Nature,</subfield>
    <subfield code="c">2023.</subfield>
  </datafield>
  <datafield tag="300" ind1=" " ind2=" ">
    <subfield code="a">1 online resource (1 article).</subfield>
  </datafield>
  <datafield tag="336" ind1=" " ind2=" ">
    <subfield code="a">text</subfield>
    <subfield code="b">txt</subfield>
    <subfield code="2">rdacontent</subfield>
  </datafield>
  <datafield tag="337" ind1=" " ind2=" ">
    <subfield code="a">computer</subfield>
    <subfield code="b">c</subfield>
    <subfield code="2">rdamedia</subfield>
  </datafield>
  <datafield tag="338" ind1=" " ind2=" ">
    <subfield code="a">online resource</subfield>
    <subfield code="b">cr</subfield>
    <subfield code="2">rdacarrier</subfield>
  </datafield>
  <datafield tag="500" ind1=" " ind2=" ">
    <subfield code="a">Publication type: Journal article.</subfield>
  </datafield>
  <datafield tag="500" ind1=" " ind2=" ">
    <subfield code="a">Linked open-access copy: published version.</subfield>
  </datafield>
  <datafield tag="506" ind1="0" ind2=" ">
    <subfield code="a">Open access; no subscription required for the linked open-access version.</subfield>
    <subfield code="f">Unrestricted online access</subfield>
    <subfield code="2">star</subfield>
  </datafield>
  <datafield tag="520" ind1="3" ind2=" ">
    <subfield code="a">Abstract Large language models (LLMs) represent a major advance in artificial intelligence (AI) research. However, the widespread use of LLMs is also coupled with significant ethical and social challenges. Previous research has pointed towards auditing as a promising governance mechanism to help ensure that AI systems are designed and deployed in ways that are ethical, legal, and technically robust. However, existing auditing procedures fail to address the governance challenges posed by LLMs, which display emergent capabilities and are adaptable to a wide range of downstream tasks. In this article, we address that gap by outlining a novel blueprint for how to audit LLMs. Specifically, we propose a three-layered approach, whereby governance audits (of technology providers that design and disseminate LLMs), model audits (of LLMs after pre-training but prior to their release), and application audits (of applications based on LLMs) complement and inform each other. We show how audits, when conducted in a structured and coordinated manner on all three levels, can be a feasible and effective mechanism for identifying and managing some of the ethical and social risks posed by LLMs. However, it is important to remain realistic about what auditing can reasonably be expected to achieve. Therefore, we discuss the limitations not only of our three-layered approach but also of the prospect of auditing LLMs at all. Ultimately, this article seeks to expand the methodological toolkit available to technology providers and policymakers who wish to analyse and evaluate LLMs from technical, ethical, and legal perspectives.</subfield>
  </datafield>
  <datafield tag="540" ind1=" " ind2=" ">
    <subfield code="a">Creative Commons Attribution (licence family recorded by the source; version not supplied in metadata).</subfield>
    <subfield code="u">https://openalex.org/W4381572755</subfield>
  </datafield>
  <datafield tag="546" ind1=" " ind2=" ">
    <subfield code="a">In English.</subfield>
  </datafield>
  <datafield tag="588" ind1="0" ind2=" ">
    <subfield code="a">Description generated from OpenAlex bibliographic metadata retrieved on 5 October 2026; full text linked from the source.</subfield>
  </datafield>
  <datafield tag="590" ind1=" " ind2=" ">
    <subfield code="a">Course allocation: B. Sc. Accounting.</subfield>
  </datafield>
  <datafield tag="650" ind1=" " ind2="4">
    <subfield code="a">large language models</subfield>
  </datafield>
  <datafield tag="650" ind1=" " ind2="4">
    <subfield code="a">artificial intelligence in auditing</subfield>
  </datafield>
  <datafield tag="650" ind1=" " ind2="4">
    <subfield code="a">AI governance</subfield>
  </datafield>
  <datafield tag="650" ind1=" " ind2="4">
    <subfield code="a">ethical risks of AI</subfield>
  </datafield>
  <datafield tag="650" ind1=" " ind2="4">
    <subfield code="a">AI regulation</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">B. Sc. Accounting</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">UMUN open access collection</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">auditing</subfield>
  </datafield>
  <datafield tag="655" ind1=" " ind2="4">
    <subfield code="a">Journal articles.</subfield>
  </datafield>
  <datafield tag="700" ind1="0" ind2=" ">
    <subfield code="a">Jonas Schuett</subfield>
    <subfield code="e">author.</subfield>
  </datafield>
  <datafield tag="700" ind1="0" ind2=" ">
    <subfield code="a">Hannah Rose Kirk</subfield>
    <subfield code="e">author.</subfield>
  </datafield>
  <datafield tag="700" ind1="0" ind2=" ">
    <subfield code="a">Luciano Floridi</subfield>
    <subfield code="e">author.</subfield>
  </datafield>
  <datafield tag="773" ind1="0" ind2=" ">
    <subfield code="t">AI and Ethics</subfield>
    <subfield code="g">Vol. 4, no. 4, (2023), pages/article 1085-1115</subfield>
    <subfield code="x">2730-5953</subfield>
  </datafield>
  <datafield tag="856" ind1="4" ind2="0">
    <subfield code="u">https://link.springer.com/content/pdf/10.1007/s43681-023-00289-2.pdf</subfield>
    <subfield code="y">Read open-access publication</subfield>
    <subfield code="z">Open access. Access and licence reported by the source; this host could not be independently checked by automation.</subfield>
    <subfield code="7">0</subfield>
  </datafield>
  <datafield tag="884" ind1=" " ind2=" ">
    <subfield code="a">Course selection and automated MARC21 enrichment from OpenAlex metadata.</subfield>
    <subfield code="g">20261005</subfield>
    <subfield code="k">https://openalex.org/W4381572755</subfield>
    <subfield code="q">local</subfield>
  </datafield>
  <datafield tag="942" ind1=" " ind2=" ">
    <subfield code="n">0</subfield>
  </datafield>
  <datafield tag="999" ind1=" " ind2=" ">
    <subfield code="c">816</subfield>
    <subfield code="d">816</subfield>
  </datafield>
</record>
