Policy and P ac ice
Liabili y Risks o Ambien Clinical Wo kflows Wi h A ificial
In elligence o Clinicians, Hospi als, and Manu ac u e s
Sa a Ge ke, Dipl-Ju Uni , MA
1
; Da id A. Simon, PhD, JD, LLM
2
; and Benjamin R. Roman, MD, MBA, MSHP
3
DOI h ps://doi.o g/10.1200/OP-24-01060
In Augus 2024, he na ion’s la ges nonp ofi in eg a ed heal h ca e p o ide , Kaise
Pe manen e, announced ha clinicians would ha e access o an ambien clinical documen a ion
sc ibe: an assis ed clinical documen a ion ool ha uses a ificial in elligence (AI) o secu ely
summa ize ele an medical in o ma ion om spoken, na u al con e sa ions (also called
ambien clinical documen a ion o AI sc ibes).
1
A e au oma ically summa izing he encoun e ,
he AI sc ibe sends he summa y o he clinician o e iew. Ambien clinical documen a ion
sc ibes a e now o e ed by some o he as es -g owing AI companies in heal h ca e, wi h
significan en u e capi al unding and an imp essi e os e o heal h sys em cus ome s.
Technologies such as ambien clinical documen a ion and o he gene a i e AI ools may im-
p o e ca e and lessen clinician bu nou by educing documen a ion bu dens. Bu hey also aise
he ques ion o who is esponsible when AI-gene a ed pa ien in o ma ion is inaccu a e, es-
pecially when hose e o s cause inju y o a pa ien . This ques ion is pa icula ly acu e in cance
ca e, whe e he e is a unique se o e minology o each o he mo e han 400 ypes o cance ,
leading o an inc eased chance o documen a ion e o , and whe e decisions on he basis o he
assump ion o in o ma ion accu acy can be li e-al e ing.
2
AI ansc ip ion ools in hei cu en e sions a e no conside ed egula ed medical de ices
unde he US Fede al Food, D ug, and Cosme ic Ac .
3
Unless his changes, he esponsibili y alls
o s akeholde s o he han he US Food and D ug Adminis a ion (FDA) o ensu e he ech-
nology’s sa e y and e ficacy. In his a icle, we analyze he AI go e nance esponsibili ies and
po en ial o liabili y o clinicians, hospi als, and manu ac u e s using AI o clinical no e-
aking and sugges se e al po en ial ways o add ess hem.
TYPES OF ERRORS
Hospi als and clinicians a e likely o ace an e ol ing landscape o AI clinical no e- aking ools
wi h powe ul capabili ies and unc ions, bu also he po en ial o making mis akes. The
cu en gold s anda d o human-w i en clinical no es o cou se also has sho comings.
3
Acco ding o a ecen es ima e, much o he con en in pa ien s’medical eco ds is dupli-
ca ed, esul ing om copy-pas ing p e ious clinical no es, po en ially leading o medical
e o s.
4
Ambien clinical documen a ion echnologies can make h ee ypes o e o s. Fi s , i may omi
in o ma ion by ailing o ecognize, ansc ibe, and mo e in o ma ion o a d a no e. Second, i
may inco ec ly documen in o ma ion, such as he w ong name o dose o a medica ion. Thi d,
i can hallucina e in o ma ion, ab ica ing new and inaccu a e in o ma ion. Al hough hese
e o s a e simila o human e o s made wi hou AI sc ibes, he mechanism o add essing hem
will be di e en . Al hough clinicians may ha e a be e chance o add essing e o s made by AI
sc ibes because hey will ha e a ansc ip o he encoun e , his emains unp o en and a he
e y leas , clinicians will ha e o adop new wo kflows.
The po en ial e o s o AI sc ibes a e non i ial. Fo example, one ansc ip ion ool has ecen ly
come unde c i icism o in en ing se e al sen ences o ex , including he pa ien ’s medical
ea men , ace, and medica ion.
4
Documen a ion e o s a e pa icula ly p oblema ic in highly
specialized fields such as oncology, whe e clinicians ely on accu a e in o ma ion abou he
pa ien ’s medical his o y, medica ion o o he ea men s, as well as ad e se e en s and
Accep ed June 20, 2025
Published Augus 1, 2025
JCO Oncol P ac 00:1-5
© 2025 by Ame ican Socie y o
Clinical Oncology
View Online
A icle
ascopubs.o g/jou nal/op | Volume nnn, Issue nnn |1
Downloaded om ascopubs.o g by 37.228.250.230 on No embe 24, 2025 om 037.228.250.230
Copy igh © 2025 Ame ican Socie y o Clinical Oncology. All igh s ese ed.
esponse o he apy o li e-and-dea h ea men decisions.
E o s in his con ex may c ea e significan liabili y isks.
HUMANS IN THE LOOP
Du ing ce ain de elopmen phases o he echnology,
manu ac u e s ha e humans in he loop o e iew accu acy
and no e quali y, including ca ying ou majo upda es o
hei AI models. Howe e , once he models each clinical
p ac ice, manu ac u e s do no use a hi d-pa y clinician
e iewe — he esponsibili y alls o he clinician e iewing
and finalizing he no e.
Fo example, AI ools o e ed by companies such as
Mic oso and Ab idge eco d and ansc ibe he en i e
pa ien -clinician con e sa ion and summa ize he ansc ip
in o a d a no e.
5
The no e is hen sen o he clinician, who
is esponsible o e iewing he d a , making changes, and
finalizing he no e.
On he one hand, such echnology may ac ually imp o e
clinical documen a ion, cap u ingmo eo be e in o -
ma ion han he clinician alone, such as ad e se e en s o
pa ien s ecei ing d ugs in oncology clinical ials. On he
o he hand, his echnology may lead o new o di e en
mis akes by clinicians. Fo example, a clinician may no
ecognize an e o in he AI-d a ed no e o a a ie y o
easons, including language/cul u al di e ences, eco ding
quali y, o simple in e p e a ion. And because clinicians
ha e high caseloads and significan pape wo k obliga ions,
hey isk au oma ion bias
6
: as hea y eliance on echnology
educes cogni i e load, physicians may be less likely o ead
e e y wo d o he d a no e be o e finalizing i . Clinicians
may be sensi i e o he isk o au oma ion bias du ing
implemen a ion and likely would no use he ool a all i
hey did no us i . Howe e , once a clinician ge s used o
using he echnology, he isk o au oma ion bias inc eases.
In he u u e, e ficiency demands may d i e hospi als and
clinicians o implemen AI clinical no e- aking ools ha
skip clinician e iew al oge he . Al hough i is unclea
whe he su ficien us in his echnology could e e d i e
adop ion o ully au oma ed documen a ion, i s heo e ical
use could exace ba e he exis ing isks o pa ien ca e.
18
LIABILITY RISKS FOR CLINICIANS, HOSPITALS,
AND MANUFACTURERS
Conside he ollowing igne e: A 68-yea -old man wi h
me as a ic non–small cell lung cance had his chemo he apy
ea men no es ansc ibed using ambien clinical docu-
men a ion. The echnology inco ec ly documen ed ha he
had ailed fi s -line and second-line he apies, when in ac
he had only ailed fi s -line he apies, and he second-line
he apies we e being discussed. The oncologis e iewed he
no e, ailed o no ice he mis ake, and finalized i in he
pa ien ’s eco d. The pa ien hen swi ched o a new on-
cologis , who elied on hese no es and ini ia ed hi d-line
he apy wi h immuno he apy p ema u ely. Ins ead o
ecei ing a mo e app op ia e second-line he apy, his
hi d-line he apy esul ed in se e e immune- ela ed
pneumoni is and hospi aliza ion. This e o dep i ed
him o he ull benefi o second-line he apyandwo s-
ened his p ognosis.
Who is likely o be held liable o he pa ien ’s ha m? The
clinicians, he hospi al, and/o he AI manu ac u e ? The
basic p inciples o o law would likely apply, bu he p ecise
scope o he liabili y isks o each ac o is de e mined by a
numbe o ac o s.
Clinician Liabili y
Clinicians a e liable when hei un easonable ac ions cause
inju y o pa ien s. Reasonableness is de e mined by he
exis ing s anda d o ca e, which can be influenced by a ious
ac o s, including he na u e o he mis ake, how di ficul i
would be o spo i , and he isk i posed. Fo example, in he
igne e case in which he fi s medical oncologis e iewed
and app o ed he AI-gene a ed no e, hey could be held
liable unde a negligence heo y o he pa ien ’s inju y i a
easonable clinician would ha e no iced he no e’s inac-
cu acy. Impo an ly, wha is easonable o he physician is
o en de e mined by wha is cus oma y in ha clinician’s
pa icula special y (which is he e medical oncology).
To de end he pa ien ’s medical malp ac ice claim, he fi s
medical oncologis could a gue ha he subsequen ea ing
oncologis was liable because a easonable clinician would
confi m in o ma ion wi h a pa ien be o e ea ing hem. In
o he wo ds, a easonable oncologis would ha e e ified
wha d ugs he pa ien was p e iously ea ed wi h be o e
p oceeding wi h hi d-line he apy. Success ully p o ing
such a de ense, howe e , would equi e he fi s medical
oncologis o in oduce e idence ha he second oncologis
knew o should ha e known he no e was inaccu a e—
pe haps h ough inconsis encies in he no es o he in o -
ma ion he fi s medical oncologis o pa ien p o ided o he
second medical oncologis di ec ly. E en i his a gumen
succeeds, he fi s oncologis could s ill be liable o some o
all o he inju y because his negligence was su ficien o
cause i .
Finally, he oncologis s migh a gue ha he cause o ha m
was he plain i ’s p eexis ing cance , no hei negligence.
While his a gumen canno de ea a pa ien ’s case by i sel ,
i can block some claims and educe damages o o he s.
19
Fo
example, some cou s equi e ha he plain i mus show
ha he p obabili y o a good ou come wi hou negligence
was g ea e han 50 pe cen . I hey can, hey a e en i led o
all damages, including om he unde lying cance . O he
ju isdic ions allow claims as long as he plain i can show
any educed eco e y bu limi damages only o he p o-
po iona e loss. Wha e e equi emen s cou s impose, he
ac ha a physician made a pa ien wo se is no a pe se ba
o a iable o claim.
8
2| © 2025 by Ame ican Socie y o Clinical Oncology
Ge ke, Simon, and Roman
Downloaded om ascopubs.o g by 37.228.250.230 on No embe 24, 2025 om 037.228.250.230
Copy igh © 2025 Ame ican Socie y o Clinical Oncology. All igh s ese ed.
Hospi al Liabili y
The hospi al whe e he fi s medical oncologis o he second
medical oncologis p ac ices may also be liable on a leas
wo heo ies. Fi s , he hospi al could be liable i he medical
oncologis s we e negligen . Unde a doc ine known as
espondea supe io , employe s a e liable o he o s o
hei employees commi ed wi hin he scope o employmen
o wi h he employe ’s p ope y.
7
I he clinicians we e
employees, he hospi al would be liable o hei negligence
since he fi s medical oncologis used he AI ool and he
second medical oncologis elied on i s no es wi hin he
scope o hei employmen .
20
Second, e en i nei he he fi s medical oncologis no he
second medical oncologis was negligen , he hospi al whe e
he fi s medical oncologis p ac ices could s ill be liable i i
ailed o mee i s du y o exe cise easonable ca e in
adop ing, implemen ing, o moni o ing he use o he AI
ool. A hospi al could b each i s du y o ca e, o example, i i
adop s an AI clinical no e- aking ool wi h a highe e o a e
han i s cu en p ac ice (ie, human no e- aking). Howe e ,
p ope ly alida ed sys ems ha educe he isk associa ed
wi h he cu en p ac ice may make he AI ool easonable o
implemen . The hospi al may also be liable i i alida ed i in
one p ac ice a ea (eg, in e nal medicine) and implemen ed i
in o he a eas (eg, oncology), o i i did no p ope ly ain
clinicians on how o use i . Fo example, i he hospi al in he
igne e desc ibed abo e did no ins uc he fi s medical
oncologis using he AI ool abou he po en ial o mis akes
o he ype o mis akes likely o a ise, he hospi al may be
liable o ha m caused by his ailu e.
Hospi als may de end hese liabili y claims by showing ha
he AI ool was p ope ly alida ed, ha he manu ac u e did
no p o ide adequa e aining be o e adop ion, o ha he
ool was de ec i e o some eason, such as no mee ing he
manu ac u e ’s own specifica ions. Howe e , any de ense
ha asse s he AI manu ac u e ’s liabili y may al eady be
limi ed by con ac law in wo un ela ed ways. Fi s , con ac
law imposes p oduc liabili y on he p oduce o a good bu
no a se ice; i a cou finds AI is a se ice, hen he AI
manu ac u e will be immune om con ac -based p oduc
liabili y sui s. Second, manu ac u e s may limi hei own
liabili y h ough con ac s wi h hospi als ha use hei
p oduc s.
8
Manu ac u e Liabili y
Manu ac u e s o AI clinical no e- aking ools could also ace
liabili y o inju ies o pa ien s in scena ios such as he ones
in he igne e unde negligence o p oduc s liabili y law.
Like hospi als, he manu ac u e can be negligen in aining
he clinicians who use he ool. Unlike hospi als and clini-
cians, howe e , manu ac u e s could be liable e en i hei
p ac ices ma ched an indus y cus om. Suppose, o exam-
ple, ha he manu ac u e hi ed a hi d-pa y fi m loca ed in
a o eign coun y ha employed non-na i e English
speake s o conduc he ini ial quali y e iew o hei ool,
be o e deploymen o clinicians. I a easonable manu ac-
u e had hi ed a na i e English speake o p o ided ap-
p op ia e aining o he fi s medical oncologis in he
igne e ha would ha e esul ed in hem ca ching he e o ,
hen he manu ac u e could be liable.
Manu ac u e s could also be liable i hei p oduc s a e
de ec i e. Failing o p o ide adequa e wa nings o make he
AI clinical no e- aking ool sa e o he in ended uses o he
p oduc , o example, may c ea e liabili y o he manu-
ac u e .
9
Wa nings mus be p o ided o consume s o he
p oduc , in his case, o he fi s oncologis —and possibly
e en o he second oncologis who e iews he comple ed
pa ien his o y and no es. Manu ac u e s who ail o disclose
isks, alida ion echniques, and o he in o ma ion could
ace liabili y i a pa ien is inju ed om hei use.
10
Fo ex-
ample, he manu ac u e could be liable i i does no in o m
he hospi al o he fi s medical oncologis using he AI o i s
isks, including he isk ha he ool may no wo k as well in
ce ain se ings o ega ding in o ma ion om ce ain
special ies, such as oncology, ha di e om hose used o
ain o alida e i .
Likewise, manu ac u e s ha do no p ope ly alida e hei
ools can ace liabili y. Because many AI clinical no e- aking
ools a e no likely o be classified as FDA- egula ed de ices,
wha will be su ficien alida ion in he igne e will depend
on ei he use (eg, heal h ca e p o ide ) expec a ions abou
he accu acy o he ool o he po en ial benefi o he ool
ela i e o i s isks, as well as whe he any al e na e sa e
designs we e a ailable.
11
MITIGATING ERRORS AND LIABILITY
Clinicians and Hospi als
Hospi als and he clinicians hey employ who use AI clinical
no e- aking ools can add ess liabili y isks in se e al ways.
Clinicians and hospi als should wo k closely wi h AI man-
u ac u e s o ensu e ha clinicians a e p ope ly ained o
use he echnology ac oss di e en ea men se ings.
T aining should help clinicians o spo common AI mis akes
o hose ha look co ec bu a e ins ead inco ec o am-
biguous. This could educe he isk ha he clinician ap-
p o es a no e ha con ains an e o . Impo an ly, aining
should no be a one-o : clinicians should be e ained when
manu ac u e s upda e hei unde lying AI models, he use
in e ace, o wo kflows ha a ec s documen a ion. Clini-
cians need o us ha he echnology is sa e and e ec i e;
his can be que ied wi h a alida ed clinician su ey, such as
he Theo y o us and accep ance o a ificial in elligence
echnology (TR AAIT).
12
Hospi als should also be me hodical when deciding on and
implemen ing AI echnologies. Mul iple gene al amewo ks
JCO Oncology P ac ice ascopubs.o g/jou nal/op | Volume nnn, Issue nnn |3
AI Clinical No e aking Risks o Oncology
Downloaded om ascopubs.o g by 37.228.250.230 on No embe 24, 2025 om 037.228.250.230
Copy igh © 2025 Ame ican Socie y o Clinical Oncology. All igh s ese ed.
exis o esponsible AI go e nance and quali y assu ance.
Recen ly, esea che s ha e de eloped one o oncology, wi h
all o i s unique clinical and ope a ional nuances.
13
Simila o
pha macy and he apeu ics commi ees o d ugs, hospi als
can es ablish commi ees o ask o ces ha a e esponsible
o AI li ecycle go e nance, egula o y compliance, and isk
managemen .
14
These bodies would be esponsible o eg-
is e ing, e alua ing, and moni o ing AI models used ac oss a
a ie y o se ings including in clinical p ac ice. They would
also be esponsible o cap u ing and e alua ing ca e quali y
issues ha a ise because o AI echnologies.
Rega ding ambien clinical documen a ion in pa icula ,
hospi als and manu ac u e s should es ablish a plan o
alida e and con inually moni o o e o s and sa e y issues
in he AI models’ou pu and finalized clinical no es ac oss
di e en clinical se ings. One esou ce o assessmen is he
Physician Documen a ion Quali y ins umen (PDQI), which
is an assessmen ool ha can be used o e alua e he quali y
o AI-gene a ed no es.
15,16
These kinds o ongoing assess-
men s may also inc ease awa eness o e o s on he pa o
clinicians who will be engaged in he e alua ion p ocess.
Howe e , no all hospi als ha e su ficien esou ces o c ea e
and main ain AI go e nance bodies. To mi iga e his p ob-
lem, hospi als wi h ewe esou ces could pa ne wi h well-
esou ced hospi als on specific AI applica ions and could
a emp o nego ia e a o able indemnifica ion p o isions in
con ac s wi h manu ac u e s.
Fo bo h legal and e hical easons, hospi als and clinicians
should be cau ious abou he po en ial u u e s a e in which
AI clinical no e- aking occu s wi hou clinician e iew. As a
legal ma e , o law liabili y o such AI ools is s ill un-
se led. As an e hical ma e , hospi als and clinicians should
conside he desi abili y o ha ing a nonhuman AI ool decide
wha in o ma ion is ele an and goes in o he medical
eco ds, especially in highly specialized fields such as on-
cology, whe e he e can be p o ound e ec s on pa ien ca e.
The e o e, main aining a human in he loop may be he mos
p uden cou se o ac ion un il o law becomes mo e se led
and he use ulness o such ools is igo ously p o en.
17
Manu ac u e s
The p ima y esponsibili y o e alua ing AI models o
clinical documen a ion should lie wi h he manu ac u e s.
They can add ess some liabili y isks in ways simila o
clinicians and hospi als. Fo ins ance, p ope ly alida ing he
echnology in specific clinical se ings is c ucial o educing
liabili y isks. The manu ac u e s should con inue o e al-
ua e and alida e he ool o ensu e i is wo king a leas as
well as when i was sold o he end use .
When human e iew is being used oe alua e model upda es,
manu ac u e s should be ca e ul in hi ing and aining hese
e iewe s. Manu ac u e s should ca e ully conside ying o
ou sou ce any human e iew unc ions o coun ies whe e
he language o he e iewe ma ches hose o he clinicians.
5
Because indemnifica ion and hold ha mless p o isions a e
only as good as hose signing hem, manu ac u e s ha do
ou sou ce such unc ions mus pay close a en ion o he
choice o law p o isions and o he financial whe ewi hal o
fi ms hey con ac wi h.
Finally, clinical documen a ion ools should only be ma -
ke ed in ways ha a e consis en wi h hei ac ual unc ions
and abili ies. Fo example, i he e is an e o a e o 20%
when used in oncology bu 2% when used in p ima y ca e,
he manu ac u e should no ma ke he p oduc as ha ing a
low e o a e. Manu ac u e s should ma ke he ool o only
hose specific uses in p ac ice a eas o se ings in which he
ool has been alida ed. Simila ly, manu ac u e s should
disclose he isks and limi a ions o he ool o ally and in
w i ing o hospi als and clinicians.
In conclusion, ambien clinical documen a ion wi h AI ools
has he po en ial o ans o m he li es o clinicians, e-
ducing bu nou and imp o ing he abili y o clinicians o
ocus on being engaged in con e sa ion wi h he pa ien
be o e hem. We ha e desc ibed po en ial issues wi h he
ools’sa e y and he ela ed liabili y conce ns. We ha e also
sugges ed a ious ways ha clinicians, hospi als, and
manu ac u e s can mi iga e hese issues. The key o en-
su ing he app op ia e use and sa e y o hese ools is he
need o clinicians, hospi als, and manu ac u e s o wo k
oge he o unde s and how his echnology is shaping and
imp o ing medical ca e. These isk- educ ion s a egies
ocus on building sa e p oduc s and p o iding comple e and
accu a e in o ma ion o clinicians and hospi als. I manu-
ac u e s, clinicians, and hospi als can sa ely implemen his
ool wi h obus AI go e nance, i may se e as a oadmap o
implemen a ion o o he AI ools such as Clinical Decision
Suppo (CDS) in clinical wo kflow. And i should imp o e
pa ien ca e.
AFFILIATIONS
1
College o Law and Eu opean Union Cen e , Uni e si y o Illinois
U bana-Champaign, Champaign, IL
2
No heas e n Uni e si y School o Law, Bos on, MA
3
Depa men o S a egy & Inno a ion and Depa men o Su ge y,
Memo ial Sloan Ke e ing Cance Cen e , New Yo k, NY
CORRESPONDING AUTHOR
Da id A. Simon, PhD, JD, LLM; e-mail: [email p o ec ed].
DISCLAIMER
The iews and opinions exp essed a e hose o he au ho s only and do
no necessa ily eflec hose o hei o ganiza ions o he g an o s— he
Eu opean Union, he Heal h and Digi al Execu i e Agency, o he
Na ional Ins i u es o Heal h. Nei he he Eu opean Union no he
4| © 2025 by Ame ican Socie y o Clinical Oncology
Ge ke, Simon, and Roman
Downloaded om ascopubs.o g by 37.228.250.230 on No embe 24, 2025 om 037.228.250.230
Copy igh © 2025 Ame ican Socie y o Clinical Oncology. All igh s ese ed.
g an ing au ho i ies can be held esponsible o hem. The unde had
no ole in p epa ing, e iewing, o app o ing he manusc ip o in
deciding o submi i o publica ion.
SUPPORT
D.S. and S.G.’s wo k was suppo ed by g an 101057321 om he
Eu opean Union. B.R.’s wo k was suppo ed by he Memo ial Sloan
Ke e ing Cance Cen e Suppo G an P30 CA008748 om he
Na ional Ins i u es o Heal h/Na ional Cance Ins i u e.
AUTHORS’DISCLOSURES OF POTENTIAL CONFLICTS
OF INTEREST
Disclosu es p o ided by he au ho s a e a ailable wi h his a icle a DOI
h ps://doi.o g/10.1200/OP-24-01060.
AUTHOR CONTRIBUTIONS
Concep ion and design: Sa a Ge ke, Da id A. Simon
Financial suppo : Sa a Ge ke
Adminis a i e suppo : All au ho s
P o ision o s udy ma e ials o pa ien s: Benjamin R. Roman
Collec ion and assembly o da a: All au ho s
Da a analysis and in e p e a ion: All au ho s
Manusc ip w i ing: All au ho s
Final app o al o manusc ip : All au ho s
Accoun able o all aspec s o he wo k: All au ho s
REFERENCES
1. Kaise Pe manen e Imp o es Membe Expe ience Wi h AI-Enabled Clinical Technology, 2024. h ps://abou .kaise pe manen e.o g/news/p ess- elease-a chi e/kaise -pe manen e-imp o es-
membe -expe ience-wi h-ai-enabled-clinical- echnology
2. NCI: Enhancing Pa ien Na iga ion wi h Technology o Imp o e Equi y in Cance Ca e
3. Chen C, Tho n on JE: AI-gene a ed clinical summa ies. JAMA 331:1967-1968, 2024
4. S einkamp J, Kan owi z JJ, Ai an-Ja ia S: P e alence and sou ces o duplica e in o ma ion in he elec onic medical eco d. JAMA Ne w Open 5:e2233348, 2022
5. T ang B: As AI p omises o e olu ionize medical no e- aking, conce ns moun abou accu acy and ha m. STAT, 2023. h ps://www.s a news.com/2023/03/28/hospi als-ai-a ificial-in elligence-
heal h-mic oso -nuance-no es/
6. Lyell D, Coie a E: Au oma ion bias and e ifica ion complexi y: A sys ema ic e iew. J Am Med In o m Assoc 24:423-431, 2017
7. Lisa M. . Hen y Mayo Newhall Meml. Hosp., 907 P.2d 358, 360 (Cal. 1995)
8. Du ou c M, Ge ke S: Decoding U.S. To Liabili y in Heal hca e’s Black-Box AI E a: Lessons om he Eu opean Union, 27 S an. Tech. L. Re . 1 (2024)
9. Simon DA: O -label p eemp ion. Wis L Re 2024:1079, 2024
10. Simon DA, Kesselheim AS: Physician and de ice manu ac u e o liabili y o emo e pa ien moni o ing de ices, in Shacha C, K ame DB, Cohen IG, e al (eds): Digi al Heal h Ca e Ou side o
T adi ional Clinical Se ings: E hical, Legal, and Regula o y Challenges and Oppo uni ies. Camb idge Uni e si y P ess, Camb idge, UK, 2024. pp 109-122
11. Mikolajczyk . Fo d Mo o Co., 901 N.E.2d 329, 352 (Ill. 2008), Opinion Modified on Denial o Reh’g (Dec. 18, 2008)
12. S e ens AF, S e son P: Theo y o us and accep ance o a ificial in elligence echnology (T AAIT): An ins umen o assess clinician us and accep ance o a ificial in elligence. J Biomed
In o m 148:104550, 2023
13. S e son PD, Choy J, Summe ille N, e al: Responsible a ificial in elligence go e nance in oncology. NPJ Digi Med 8:407, 2025
14. Simon DA: O -label inno a ion. Ga L Re 56:701-786, 2022
15. S e son PD, Bakken S, W enn JO, e al: Assessing elec onic no e quali y using he physician documen a ion quali y ins umen (PDQI-9). Appl Clin In o m 3:164-174, 2012
16. C ox o d E, Gao Y, Pelleg ino N, e al: De elopmen and alida ion o he p o ide documen a ion summa iza ion quali y ins umen o la ge language models. J Am Med In o m Assoc 32:
1050-1060, 2025
17. Cahill RA, Du ou c M, Ge ke S: Su gical ideo da a—“In,”“ou ,”o “shake i all abou ” he medical eco d. Ann Su g 281:382-384, 2025
18. Simon DA: Pa ien s e sus p ofi s. Vand J En Tech L 2025. h ps://pape s.ss n.com/sol3/pape s.c m?abs ac _id55296927
19. Lo d . Lo e , 770 A.2d 1103 (N.H. 2001).
20. Simon DA, Cohen IG, Bala ba C, e al: The hospi al-a -home p esen s no el liabili ies o physicians, hospi als, ca egi e s, and pa ien s. Na Med 28:438-441, 2022
JCO Oncology P ac ice ascopubs.o g/jou nal/op | Volume nnn, Issue nnn |5
AI Clinical No e aking Risks o Oncology
Downloaded om ascopubs.o g by 37.228.250.230 on No embe 24, 2025 om 037.228.250.230
Copy igh © 2025 Ame ican Socie y o Clinical Oncology. All igh s ese ed.
AUTHORS’DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
Liabili y Risks o Ambien Clinical Wo kflows Wi h A ificial In elligence o Clinicians, Hospi als, and Manu ac u e s
The ollowing ep esen s disclosu e in o ma ion p o ided by au ho s o his manusc ip . All ela ionships a e conside ed compensa ed unless
o he wise no ed. Rela ionships a e sel -held unless no ed. I 5Immedia e Family Membe , Ins 5My Ins i u ion. Rela ionships may no ela e o he
subjec ma e o his manusc ip . Fo mo e in o ma ion abou ASCO’s conflic o in e es policy, please e e o www.asco.o g/ wc o
ascopubs.o g/op/au ho s/au ho -cen e .
Open Paymen s is a public da abase con aining in o ma ion epo ed by companies abou paymen s made o US-licensed physicians (Open
Paymen s).
Da id A. Simon
S ock and O he Owne ship In e es s: Pfize , P oc e & Gamble
Hono a ia: Chan Zucke be g Ini ia i e, Geo ge Mason Uni e si y
O he Rela ionship: A nold Ven u es, Simon Law G oup
No o he po en ial conflic s o in e es we e epo ed.
© 2025 by Ame ican Socie y o Clinical Oncology
Ge ke, Simon, and Roman
Downloaded om ascopubs.o g by 37.228.250.230 on No embe 24, 2025 om 037.228.250.230
Copy igh © 2025 Ame ican Socie y o Clinical Oncology. All igh s ese ed.
