Frontiers in Public Health,
Journal Year:
2025,
Volume and Issue:
13
Published: May 9, 2025
Introduction
Generative
artificial
intelligence
(AI)
is
advancing
rapidly;
an
important
consideration
the
public’s
increasing
ability
to
customise
foundational
AI
models
create
publicly
accessible
applications
tailored
for
specific
tasks.
This
study
aims
evaluate
accessibility
and
functionality
descriptions
of
customised
GPTs
on
OpenAI
GPT
store
that
provide
health-related
information
or
assistance
patients
healthcare
professionals.
Methods
We
conducted
a
cross-sectional
observational
from
September
2
6,
2024,
identify
with
functions.
searched
across
general
medicine,
psychology,
oncology,
cardiology,
immunology
applications.
Identified
were
assessed
their
name,
description,
intended
audience,
usage.
Regulatory
status
was
checked
U.S.
Food
Drug
Administration
(FDA),
European
Union
Medical
Device
Regulation
(EU
MDR),
Australian
Therapeutic
Goods
(TGA)
databases.
Results
A
total
1,055
customised,
targeting
professionals
identified,
which
had
collectively
been
used
in
over
360,000
conversations.
Of
these,
587
psychology-related,
247
105
52
30
immunology,
34
other
health
specialties.
Notably,
624
identified
included
professional
titles
(e.g.,
doctor,
nurse,
psychiatrist,
oncologist)
names
and/or
descriptions,
suggesting
they
taking
such
roles.
None
FDA,
EU
MDR,
TGA-approved.
Discussion
highlights
rapid
emergence
accessible,
GPTs.
The
findings
raise
questions
about
whether
current
medical
device
regulations
are
keeping
pace
technological
advancements.
results
also
highlight
potential
“role
creep”
chatbots,
where
begin
perform
—
claim
functions
traditionally
reserved
licensed
professionals,
underscoring
safety
concerns.
JAMA Network Open,
Journal Year:
2025,
Volume and Issue:
8(4), P. e258052 - e258052
Published: April 30, 2025
Importance
The
primary
objective
of
any
newly
developed
medical
device
using
artificial
intelligence
(AI)
is
to
ensure
its
safe
and
effective
use
in
broader
clinical
practice.
Objective
To
evaluate
key
characteristics
AI-enabled
devices
approved
by
the
US
Food
Drug
Administration
(FDA)
that
are
relevant
their
generalizability
reported
public
domain.
Design,
Setting,
Participants
This
cross-sectional
study
collected
information
on
all
received
FDA
approval
were
listed
website
as
August
31,
2024.
Main
Outcomes
Measures
For
each
device,
detailed
for
at
time
summarized,
specifically
examining
evaluation
aspects,
such
presence
design
performance
studies,
availability
discriminatory
metrics,
age-
sex-specific
data.
Results
In
total,
903
FDA-approved
analyzed,
most
which
became
available
last
decade.
primarily
related
specialties
radiology
(692
[76.6.%]),
cardiovascular
medicine
(91
[10.1%]),
neurology
(29
[3.2%]).
Most
software
only
(664
[73.5%]),
6
(0.7%)
implantable.
Detailed
descriptions
development
absent
from
publicly
provided
summaries.
Clinical
studies
505
(55.9%),
while
218
(24.1%)
explicitly
stated
no
conducted.
Retrospective
designs
common
(193
[38.2%]),
with
41
(8.1%)
being
prospective
12
(2.4%)
randomized.
Discriminatory
metrics
200
summaries
(sensitivity:
183
[36.2%];
specificity:
176
[34.9%];
area
under
curve:
82
[16.2%]).
Among
less
than
one-third
data
(145
[28.7%]),
117
(23.2%)
addressed
age-related
subgroups.
Conclusions
Relevance
this
study,
approximately
half
devices,
yet
was
often
insufficient
a
comprehensive
assessment
generalizability,
emphasizing
need
ongoing
monitoring
regular
re-evaluation
identify
address
unexpected
changes
during
use.
Frontiers in Public Health,
Journal Year:
2025,
Volume and Issue:
13
Published: May 9, 2025
Introduction
Generative
artificial
intelligence
(AI)
is
advancing
rapidly;
an
important
consideration
the
public’s
increasing
ability
to
customise
foundational
AI
models
create
publicly
accessible
applications
tailored
for
specific
tasks.
This
study
aims
evaluate
accessibility
and
functionality
descriptions
of
customised
GPTs
on
OpenAI
GPT
store
that
provide
health-related
information
or
assistance
patients
healthcare
professionals.
Methods
We
conducted
a
cross-sectional
observational
from
September
2
6,
2024,
identify
with
functions.
searched
across
general
medicine,
psychology,
oncology,
cardiology,
immunology
applications.
Identified
were
assessed
their
name,
description,
intended
audience,
usage.
Regulatory
status
was
checked
U.S.
Food
Drug
Administration
(FDA),
European
Union
Medical
Device
Regulation
(EU
MDR),
Australian
Therapeutic
Goods
(TGA)
databases.
Results
A
total
1,055
customised,
targeting
professionals
identified,
which
had
collectively
been
used
in
over
360,000
conversations.
Of
these,
587
psychology-related,
247
105
52
30
immunology,
34
other
health
specialties.
Notably,
624
identified
included
professional
titles
(e.g.,
doctor,
nurse,
psychiatrist,
oncologist)
names
and/or
descriptions,
suggesting
they
taking
such
roles.
None
FDA,
EU
MDR,
TGA-approved.
Discussion
highlights
rapid
emergence
accessible,
GPTs.
The
findings
raise
questions
about
whether
current
medical
device
regulations
are
keeping
pace
technological
advancements.
results
also
highlight
potential
“role
creep”
chatbots,
where
begin
perform
—
claim
functions
traditionally
reserved
licensed
professionals,
underscoring
safety
concerns.
