May 30, 2025 • 12 mins
Health Affairs' Jeff Byers welcomes Senior Editor Michael Gerber back to the program to discuss the Food and Drug Administration's recent announcement to scale a generative artificial intelligence across its center in the future.
Health Affairs released their first Insider trend report. The report focuses on AI in health care and you can get full access to this report by becoming an Insider. Insiders also will receive access to our June 17 event on risk adjustment trends and our July 7 event featuring a wrap-up of the recent Supreme Court session.
Related Links:
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Jeff Byers (00:08):
Hello and welcome to Health Affairs This Week. I'm
your host Jeff Beyers. We'rerecording on 05/28/2025. Join
our insider program and gainaccess to our upcoming summer
slam of health policy. On June17, we have a risk adjustments
primer event with J.
Michael McWilliams. And on July9, Sarah Rosenbaum will moderate
(00:29):
a panel wrapping up the currentsupreme court session. This is
the fourth time we've hostedthis popular event, so join
Insider to join us. And today onthe pod, we have health affairs
own Michael Gerber. Last week,Insider released its first trend
report on AI and health care.
This is hot on the heels of theFDA releasing a timeline this
(00:50):
month to scale AI across itscenters. While we won't dive
into the actual report, you'llhave to join Insider for that,
we wanted to talk about FDA's AIefforts, what it says, what it
doesn't say, etcetera, etcetera.Michael, welcome back to the
program.
Michael Gerber (01:05):
Hey, Jeff. Thanks a lot for having me.
Jeff Byers (01:07):
Michael, so how would you define your
relationship with technology?
Michael Gerber (01:12):
That is a great question. I mean, I was able to,
you know, log in to record thispodcast, so I feel like that was
a success. But I am certainlynot an early adopter when it
comes to AI and would probablysay sometimes sometimes a
skeptic, although I'm trying tobe open minded about it and
certainly see the benefits itcan have.
Jeff Byers (01:34):
Yeah. So when I say the words or phrase single sign
on, do you get filled with dreador excitement?
Michael Gerber (01:40):
I I mostly I think I can handle single sign
on, Jeff. But beyond that, itstarts to get a little tricky.
Jeff Byers (01:47):
Yeah. When looking to AI, we are men of a certain
age. I'm probably in the sameboat as you are where I'm
optimistically skeptical aboutthe phrase and technology
itself. Because as we talkedabout off mic, it can be used to
be a very broad descriptor or avery specific pinpoint
descriptor. And with thatframing, as we talk about the
(02:11):
FDA press release, I think itkinda lends itself to the to the
vague descriptors.
That's just my interpretation ofit. So looking at the FDA
release, they released atimeline to try and scale AI
across their centers. By theJune, which checking on my
(02:31):
calendar is within thirty daysor so, so not a long time. This
follows the completion of agenerative AI pilot. So what can
you tell us about the timelineand the general purpose of this
Gen AI use at the FDA?
Michael Gerber (02:44):
Yeah. I mean, based on what the FDA said and
and the minimal other reportingout there, there's not a lot
more to really say about it.They they did release this
announcement where, you know,the new FDA commissioner said he
was, quote, blown away by thesuccess of their first AI
assisted scientific reviewpilot, but there was no more
(03:07):
detail really as to what thatmeans, what what that pilot was,
how extensive it was. Was itused just for one piece of the
review of one, you know, productor drug? Was it used more widely
than that?
And at the same time, you know,he he also said that they have
an aggressive timeline to scaleuse of AI internally across all
(03:29):
FDA centers, so, you know, allthe centers offices within the
FDA by June 30, which as yousaid is just a month away. But
it's also unclear on what thatmeans. I mean, if every center
is using it, but how much arethey using it? Because they say
full integration by the June,which sounds pretty extensive,
but also that they will continueto refine it after. I mean, they
(03:52):
say, quote, to expand use cases,improve functionality, and adapt
to the evolving needs.
So they're not really sayingevery use case will be
determined by the next month.But certainly excited about
using AI to review, you know,different drugs, different
medical devices, food, whateverelse the FDA is doing to to do
(04:12):
those reviews faster and moreefficiently.
Jeff Byers (04:14):
Yeah. Thanks thanks for that quick overview. Yeah.
It it's a little vague, alsothere has been some news reports
on it, not a large amount. StatNews published a under paywall
article called, like, three orfive burning questions about
FDA's AI review.
So there are some people lookinginto this. What were some of the
details not included?
Michael Gerber (04:34):
Right. Well well, as as we sort of alluded
to, some of the details areexactly how they're using this.
So and and also how how muchthis announcement actually means
anything new is happening. It'snot as if the FDA wasn't looking
at AI and using some form of AIalready. I mean, even defining
what AI is, right, as we talkedabout is tough, you know, to
(04:56):
some of us have sort of beenusing AI since Google was
created in some ways, and it'sit's a sliding scale.
But some of the things weren'ttalked about is, you know,
there's some obvious sort ofsimple, I think, use cases that
maybe even to you and I don'tsound that scary. When a company
or manufacturer is is submittingfor approval at the FDA, you
(05:19):
know, it's it's hundreds, if notthousands of pages of paperwork
often of scientific data andother information they have to
submit. And maybe AI is alreadyor can be used to just sort of
look at that application and seeif anything's missing. And
without a human having to dothat first step, which can be
very tedious and time consumingjust to go back to someone and
(05:41):
say, hey, you forgot this pageor this isn't filled out right.
So that's sort of the first stepin how AI might be used.
But then, you know, all the wayup to actually evaluating the
data and comparing it toprevious applications and other,
you know, drugs, for example, tosee the safety and effectiveness
of the drug and and help thescientists, you know, go through
(06:04):
all that and determine exactlywhat they're looking at.
Jeff Byers (06:08):
You know, this press release kinda reminds me of my
time as a business reporter whenI was working at Health Care
Dive where you'd get so many, somany press releases. And some of
them had some really interestinginformation, and some of them
you could kinda tell were justcreated to, like, people that a
company actually existed. So Iyou know, FDA is a lot different
(06:31):
in that that they are not tryingto, like, get venture funds for
the most part, but it was alittle vague, and maybe they're
trying to get some shine thatcomes from AI or talking about
AI. And, you know, we at HealthAffairs are trying to promote
our AI and health care insiderreport, which you can get for
signing up for insider for ahundred and $80 a year. And so
(06:54):
based on that, now we're talkingabout it, the FDA release.
So interesting with that. Goingback to the press release,
again, kinda going on what wesee in it, does this signify a
shift in governmental thinkingregarding AI deployment?
Michael Gerber (07:12):
Yeah. I think I think yes and no. I think the
FDA announcement may or may not.I mean, they're in the the
little reporting that hashappened since the announcement,
you know, Robert Califf, theformer commissioner under
president Biden and and previousto that as well, you know, said
we were already using AI. Thiswas already underway, but he did
(07:33):
express a little bit, you know,a little concern over this
timeline.
Again, not knowing exactly whatit means, but that they don't
wanna rush into it because we doknow the dangers of using AI in
different ways, and I thinkthat's, you know, where you and
I have certainly seen that someof the skepticism we have, and
within healthcare and algorithmsand things like that. So I think
it might be a shift in in justdoing it faster at the FDA, but
(07:57):
then we don't know again exactlywhat it means. I think that the
Biden administration certainly,you know, one of did put out an
executive order that proposedsome sort of, you know,
guardrails or potentialregulations, and and they were
not not necessarily anti AI, butcertainly trying to be careful
and think about it. And, youknow, beyond the FDA or even
(08:18):
beyond health care, the Trumpadministration has rolled back
that executive order and just ingeneral seem to be a little more
aggressive with saying they'regonna use AI. I think the other
aspect of that, you know,commissioner McCary was pretty
careful in the press release tosay, you know, this is to help
people do their jobs better, notto replace people.
At the same time, as we know, alot of people have been let go
(08:42):
from the FDA and otherorganizations throughout the
administration. They certainlythere certainly is a fear, I
think, among people inorganizations that are regulated
by the FDA or have products upthat things would take longer
because they had fewer people.So perhaps it's a a reaction to
that as well to say, hey. Look.You know, we we might have cut
(09:02):
some of these people, but we'remaking progress with AI, which
will make things more efficientand faster.
Sort of a you know, they theydon't mention Doge directly, and
I don't think this isnecessarily straight out of
that. But it's along those samelines of, you know, maybe fewer
people, but faster computers,they can do their job just as
well, if not better.
Jeff Byers (09:21):
Yeah. So it sounds like they're pointing into a
direction of where they'regoing, but without the concrete
details, you you and I and otherpeople can probably put our
thoughts and feelings about AIonto this press release about
whether it's and how it meanwhat it means to them.
Michael Gerber (09:39):
Right. Now the the the one thing the one
reporting sort of separate fromthis, but recently, Wired
magazine did report that the FDAand their new chief AI officer,
who they recently appointed fromindustry, has a lot of
experience in consulting andtechnology and government, but
has been meeting with OpenAIabout, you know, developing a
(10:02):
specific chat GPT model, Iguess, or OpenAI model for the
FDA to use. At the same time,you know, some of the big
concerns with AI and the FDAthat people are raising are, you
know, what does it mean when youwork with those outside
companies, and will the AIsoftware or the AI model itself
be completely closed off toprevent, you know, proprietary
(10:25):
information from being leaked orbeing used to teach a model and
then being out there in theworld. So
Jeff Byers (10:31):
So what can we expect moving forward? Like,
what are the immediate nextsteps?
Michael Gerber (10:35):
Right. Well, I think the the first thing is
that the FDA did say that inJune, so that's, you know, just
a few days away, that additionaldetails and updates on the
initiative will be sharedpublicly. So we'll wait and see
if this really was just a teaserfor a lot more details or if
(10:55):
sort of another example ofsaying we're gonna tell you a
lot more and then, we only get alittle taste or not much more in
June. So, beyond that, I thinkwaiting to see really what some
of the the people who work withthe FDA, what they say. Are they
seeing reviews come back with,you know, evidence?
Are they being told AI was usedas part of this review? Are
(11:16):
reviews happening faster? Andthose things will take time.
Right? Reviews can take up to ayear, so it'll take time to see
if they're faster, if they'remore efficient, and then even
longer to see if they're evenmore safe and effective than
what happens now or not.
Jeff Byers (11:30):
So potentially lots more details to come. Many
questions to weigh and to ask inthe future. So Michael Gerber,
thanks for joining us today onHealth Affairs this week. And if
you, the listener, have morequestions about AI, you can join
Health Affairs Insider for ahundred and $80 a year and get
our new trend report on AI inhealth care, which breaks down,
(11:52):
some current uses of AI, whatare some state laws happening in
AI, and where they intersect inthe health care landscape. And
with that, share this episodewith the, generative robot in
your life, and we'll see younext week.
Thanks all.
Hello and welcome to Health Affairs This Week. I'm
your host Jeff Beyers. We'rerecording on 05/28/2025. Join
our insider program and gainaccess to our upcoming summer
slam of health policy. On June17, we have a risk adjustments
primer event with J.
Michael McWilliams. And on July9, Sarah Rosenbaum will moderate
(00:29):
a panel wrapping up the currentsupreme court session. This is
the fourth time we've hostedthis popular event, so join
Insider to join us. And today onthe pod, we have health affairs
own Michael Gerber. Last week,Insider released its first trend
report on AI and health care.
This is hot on the heels of theFDA releasing a timeline this
(00:50):
month to scale AI across itscenters. While we won't dive
into the actual report, you'llhave to join Insider for that,
we wanted to talk about FDA's AIefforts, what it says, what it
doesn't say, etcetera, etcetera.Michael, welcome back to the
program.
Michael Gerber (01:05):
Hey, Jeff. Thanks a lot for having me.
Jeff Byers (01:07):
Michael, so how would you define your
relationship with technology?
Michael Gerber (01:12):
That is a great question. I mean, I was able to,
you know, log in to record thispodcast, so I feel like that was
a success. But I am certainlynot an early adopter when it
comes to AI and would probablysay sometimes sometimes a
skeptic, although I'm trying tobe open minded about it and
certainly see the benefits itcan have.
Jeff Byers (01:34):
Yeah. So when I say the words or phrase single sign
on, do you get filled with dreador excitement?
Michael Gerber (01:40):
I I mostly I think I can handle single sign
on, Jeff. But beyond that, itstarts to get a little tricky.
Jeff Byers (01:47):
Yeah. When looking to AI, we are men of a certain
age. I'm probably in the sameboat as you are where I'm
optimistically skeptical aboutthe phrase and technology
itself. Because as we talkedabout off mic, it can be used to
be a very broad descriptor or avery specific pinpoint
descriptor. And with thatframing, as we talk about the
(02:11):
FDA press release, I think itkinda lends itself to the to the
vague descriptors.
That's just my interpretation ofit. So looking at the FDA
release, they released atimeline to try and scale AI
across their centers. By theJune, which checking on my
(02:31):
calendar is within thirty daysor so, so not a long time. This
follows the completion of agenerative AI pilot. So what can
you tell us about the timelineand the general purpose of this
Gen AI use at the FDA?
Michael Gerber (02:44):
Yeah. I mean, based on what the FDA said and
and the minimal other reportingout there, there's not a lot
more to really say about it.They they did release this
announcement where, you know,the new FDA commissioner said he
was, quote, blown away by thesuccess of their first AI
assisted scientific reviewpilot, but there was no more
(03:07):
detail really as to what thatmeans, what what that pilot was,
how extensive it was. Was itused just for one piece of the
review of one, you know, productor drug? Was it used more widely
than that?
And at the same time, you know,he he also said that they have
an aggressive timeline to scaleuse of AI internally across all
(03:29):
FDA centers, so, you know, allthe centers offices within the
FDA by June 30, which as yousaid is just a month away. But
it's also unclear on what thatmeans. I mean, if every center
is using it, but how much arethey using it? Because they say
full integration by the June,which sounds pretty extensive,
but also that they will continueto refine it after. I mean, they
(03:52):
say, quote, to expand use cases,improve functionality, and adapt
to the evolving needs.
So they're not really sayingevery use case will be
determined by the next month.But certainly excited about
using AI to review, you know,different drugs, different
medical devices, food, whateverelse the FDA is doing to to do
(04:12):
those reviews faster and moreefficiently.
Jeff Byers (04:14):
Yeah. Thanks thanks for that quick overview. Yeah.
It it's a little vague, alsothere has been some news reports
on it, not a large amount. StatNews published a under paywall
article called, like, three orfive burning questions about
FDA's AI review.
So there are some people lookinginto this. What were some of the
details not included?
Michael Gerber (04:34):
Right. Well well, as as we sort of alluded
to, some of the details areexactly how they're using this.
So and and also how how muchthis announcement actually means
anything new is happening. It'snot as if the FDA wasn't looking
at AI and using some form of AIalready. I mean, even defining
what AI is, right, as we talkedabout is tough, you know, to
(04:56):
some of us have sort of beenusing AI since Google was
created in some ways, and it'sit's a sliding scale.
But some of the things weren'ttalked about is, you know,
there's some obvious sort ofsimple, I think, use cases that
maybe even to you and I don'tsound that scary. When a company
or manufacturer is is submittingfor approval at the FDA, you
(05:19):
know, it's it's hundreds, if notthousands of pages of paperwork
often of scientific data andother information they have to
submit. And maybe AI is alreadyor can be used to just sort of
look at that application and seeif anything's missing. And
without a human having to dothat first step, which can be
very tedious and time consumingjust to go back to someone and
(05:41):
say, hey, you forgot this pageor this isn't filled out right.
So that's sort of the first stepin how AI might be used.
But then, you know, all the wayup to actually evaluating the
data and comparing it toprevious applications and other,
you know, drugs, for example, tosee the safety and effectiveness
of the drug and and help thescientists, you know, go through
(06:04):
all that and determine exactlywhat they're looking at.
Jeff Byers (06:08):
You know, this press release kinda reminds me of my
time as a business reporter whenI was working at Health Care
Dive where you'd get so many, somany press releases. And some of
them had some really interestinginformation, and some of them
you could kinda tell were justcreated to, like, people that a
company actually existed. So Iyou know, FDA is a lot different
(06:31):
in that that they are not tryingto, like, get venture funds for
the most part, but it was alittle vague, and maybe they're
trying to get some shine thatcomes from AI or talking about
AI. And, you know, we at HealthAffairs are trying to promote
our AI and health care insiderreport, which you can get for
signing up for insider for ahundred and $80 a year. And so
(06:54):
based on that, now we're talkingabout it, the FDA release.
So interesting with that. Goingback to the press release,
again, kinda going on what wesee in it, does this signify a
shift in governmental thinkingregarding AI deployment?
Michael Gerber (07:12):
Yeah. I think I think yes and no. I think the
FDA announcement may or may not.I mean, they're in the the
little reporting that hashappened since the announcement,
you know, Robert Califf, theformer commissioner under
president Biden and and previousto that as well, you know, said
we were already using AI. Thiswas already underway, but he did
(07:33):
express a little bit, you know,a little concern over this
timeline.
Again, not knowing exactly whatit means, but that they don't
wanna rush into it because we doknow the dangers of using AI in
different ways, and I thinkthat's, you know, where you and
I have certainly seen that someof the skepticism we have, and
within healthcare and algorithmsand things like that. So I think
it might be a shift in in justdoing it faster at the FDA, but
(07:57):
then we don't know again exactlywhat it means. I think that the
Biden administration certainly,you know, one of did put out an
executive order that proposedsome sort of, you know,
guardrails or potentialregulations, and and they were
not not necessarily anti AI, butcertainly trying to be careful
and think about it. And, youknow, beyond the FDA or even
(08:18):
beyond health care, the Trumpadministration has rolled back
that executive order and just ingeneral seem to be a little more
aggressive with saying they'regonna use AI. I think the other
aspect of that, you know,commissioner McCary was pretty
careful in the press release tosay, you know, this is to help
people do their jobs better, notto replace people.
At the same time, as we know, alot of people have been let go
(08:42):
from the FDA and otherorganizations throughout the
administration. They certainlythere certainly is a fear, I
think, among people inorganizations that are regulated
by the FDA or have products upthat things would take longer
because they had fewer people.So perhaps it's a a reaction to
that as well to say, hey. Look.You know, we we might have cut
(09:02):
some of these people, but we'remaking progress with AI, which
will make things more efficientand faster.
Sort of a you know, they theydon't mention Doge directly, and
I don't think this isnecessarily straight out of
that. But it's along those samelines of, you know, maybe fewer
people, but faster computers,they can do their job just as
well, if not better.
Jeff Byers (09:21):
Yeah. So it sounds like they're pointing into a
direction of where they'regoing, but without the concrete
details, you you and I and otherpeople can probably put our
thoughts and feelings about AIonto this press release about
whether it's and how it meanwhat it means to them.
Michael Gerber (09:39):
Right. Now the the the one thing the one
reporting sort of separate fromthis, but recently, Wired
magazine did report that the FDAand their new chief AI officer,
who they recently appointed fromindustry, has a lot of
experience in consulting andtechnology and government, but
has been meeting with OpenAIabout, you know, developing a
(10:02):
specific chat GPT model, Iguess, or OpenAI model for the
FDA to use. At the same time,you know, some of the big
concerns with AI and the FDAthat people are raising are, you
know, what does it mean when youwork with those outside
companies, and will the AIsoftware or the AI model itself
be completely closed off toprevent, you know, proprietary
(10:25):
information from being leaked orbeing used to teach a model and
then being out there in theworld. So
Jeff Byers (10:31):
So what can we expect moving forward? Like,
what are the immediate nextsteps?
Michael Gerber (10:35):
Right. Well, I think the the first thing is
that the FDA did say that inJune, so that's, you know, just
a few days away, that additionaldetails and updates on the
initiative will be sharedpublicly. So we'll wait and see
if this really was just a teaserfor a lot more details or if
(10:55):
sort of another example ofsaying we're gonna tell you a
lot more and then, we only get alittle taste or not much more in
June. So, beyond that, I thinkwaiting to see really what some
of the the people who work withthe FDA, what they say. Are they
seeing reviews come back with,you know, evidence?
Are they being told AI was usedas part of this review? Are
(11:16):
reviews happening faster? Andthose things will take time.
Right? Reviews can take up to ayear, so it'll take time to see
if they're faster, if they'remore efficient, and then even
longer to see if they're evenmore safe and effective than
what happens now or not.
Jeff Byers (11:30):
So potentially lots more details to come. Many
questions to weigh and to ask inthe future. So Michael Gerber,
thanks for joining us today onHealth Affairs this week. And if
you, the listener, have morequestions about AI, you can join
Health Affairs Insider for ahundred and $80 a year and get
our new trend report on AI inhealth care, which breaks down,
(11:52):
some current uses of AI, whatare some state laws happening in
AI, and where they intersect inthe health care landscape. And
with that, share this episodewith the, generative robot in
your life, and we'll see younext week.
Thanks all.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!