In
yet
another
case
involving
the
use
—
and
misuse
—
of
generative
AI
in
legal
research,
a
federal
district
court
judge
has
declined
to
impose
sanctions
on
an
experienced
attorney
who
submitted
a
brief
containing
miscited
and
misquoted
cases
generated
by
artificial
intelligence.
Although
the
attorney,
Thad
M.
Guyer,
acknowledged
using
gen
AI
tools
in
preparing
his
appellate
brief,
which
contained
several
incorrectly
cited
cases,
U.S.
District
Judge
Thomas
T.
Cullen,
presiding
over
the
whistleblower
lawsuit
in
the
Western
District
of
Virginia,
characterized
the
citation
errors
as
an
“honest
mistake”
rather
than
intentional
misconduct.
While
the
judge
in
the
case,
Iovino
v.
Michael
Stapleton
Associates,
did
not
issue
a
written
opinion
on
the
sanctions
issue,
he
made
his
ruling
orally
during
an
Oct.
9,
2024,
show
cause
hearing
held
to
determine
whether
Guyer
should
be
sanctioned.
You
can
read
the
transcript
of
the
hearing
here.
As
far
as
I
could
find,
the
order
has
not
yet
been
reported
in
any
other
news
media.
“Mr.
Guyer,
to
his
credit,
owned
the
mistake,”
Judge
Cullen
said
during
the
hearing.
“He
took
sole
responsibility,
didn’t
try
to
blame
[other]
counsel
involved,
and
immediately
took
steps
to
correct
the
issues
that
led
to
the
erroneous
case
cites.
…
This
was
a
quirk.
It’s
one
of
the
downsides
of
generative
AI.”
Erroneous
Cites
and
Quotes
The
Oregon-based
Guyer,
a
nationally
known
whistleblower
lawyer
with
his
own
Wikipedia
page,
was
representing
the
plaintiff
in
the
Virginia
lawsuit
pro
hac
vice
through
his
association
with
the
Government
Accountability
Project
(GAP),
a
whistleblower
rights
organization
for
which
he
served
earlier
in
his
career
as
litigation
director
and
general
counsel.
Last
July,
Judge
Cullen
issued
an
opinion
and
order
directing
Guyer
to
show
cause
why
he
should
not
be
disciplined,
after
it
was
discovered
that
a
brief
Guyer
had
filed,
objecting
to
a
protective
order
in
a
discovery
dispute,
contained
cites
that
could
not
be
found
and
cases
that
appeared
not
to
exist.
The
brief
also
contained
case
quotations
that
did
not
exist
in
the
cited
cases.
Opposing
counsel
flagged
these
discrepancies
to
the
court,
calling
them
“ChatGPT
run
amok.”
At
the
time
the
judge
issued
the
show
cause
order,
Guyer
had
responded
to
the
erroneous
citations
by
filing
supplemental
authorities.
However,
Judge
Cullen
wrote
in
the
order
that
Guyer
“puzzlingly”
had
not
explained
where
the
citations
had
come
from
or
who
was
to
blame.
“This
silence
is
deafening,”
the
judge
wrote.
After
the
show
cause
order
was
issued,
Guyer
filed
a
response
in
which
he
denied
using
any
fictitious
cases
but
acknowledged
using
several
erroneous
citations
and
quotations.
He
asserted
that
the
two
cases
cited
by
the
court
as
non-existent
did,
in
fact,
exist,
but
were
given
the
wrong
citations.
As
to
the
quotations,
he
said
that
while
they
did
not
appear
verbatim
in
the
cases,
they
accurately
reflected
principles
discussed
in
those
cases.
In
a
sworn
declaration
filed
with
the
court,
Guyer
described
himself
as
an
early
and
regular
user
of
gen
AI.
“I
study,
practice
and
use
Generative
AI
and
GPTs
(“GPTs”)
extensively
in
my
law
practice,
and
on
this
case,”
he
wrote.
“At
age
74,
and
being
an
early
adopter
and
proponent
of
technologies
to
aid
lawyers,
I
am
fully
committed
to
the
responsible
use
of
the
GPTs
in
this
profound
professional
transformation
of
information
technologies
for
legal
research,
document
preparation,
and
discovery.”
‘The
New
Normal’
At
the
show
cause
hearing
in
October,
the
judge
indicated
that
he,
in
general,
was
open
to
attorneys’
use
of
AI,
calling
it
“the
new
normal.”
“This
court
has
neither
the
authority
nor
the
inclination
to
curb
that
practice,
even
it
wanted
to,”
he
said.
“If
I
did
that,
I
would
justifiably
be
perceived
by
some
as
overstepping
and
unwisely
decreeing
that
litigants
can’t
use
this
groundbreaking
technology
in
a
court
of
law.”
At
the
same
time,
he
said,
litigants
who
use
gen
AI
to
prepare
pleadings
and
briefs
“must
still
adhere
to
basic
tenets
of
conduct,
including
taking
reasonable
measures
to
ensure
that
what
they
do
file
in
court,
including
cases
cited
to
bolster
legal
arguments,
is
true
and
accurate
to
the
best
of
their
ability.”
The
judge
credited
Guyer
for
accepting
sole
responsibility
for
his
error
and
for
acknowledging
that
he
did
not
employ
the
safeguards
necessary
to
ensure
that
AI
was
used
properly.
He
also
credited
Guyer
for
pointing
out
additional
citation
errors
that
both
the
court
and
opposing
counsel
had
missed.
The
judge
also
noted
that
Guyer
and
GAP
had
pledged
to
implement
safeguards
going
forward,
including
using
databases
such
as
Westlaw
to
check
AI-generated
case
cites
and
discussions
and
assigning
a
junior
attorney
to
manually
check
briefs
written
using
gen
AI.
“I
don’t
believe
that
Mr.
Guyer
did
anything
intentionally
to
mislead
this
court,”
Judge
Cullen
wrote.
“He
has
an
unblemished
track
record
as
a
lawyer,
practicing
at
a
very
high
level
over
the
course
of
his
career.
He
made
a
mistake,
and
for
the
most
part,
he’s
acknowledged
that.”
That
said,
the
judge
was
not
ready
to
let
Guyer
entirely
off
the
hook.
He
expressed
consternation
over
comments
Guyer
made
in
the
media
asserting
that
the
cases
were
real
and
that,
even
if
the
citations
were
wrong,
the
court
should
have
been
able
to
figure
that
out,
or
his
opposing
counsel
should
have
alerted
him
in
time
to
quickly
correct
the
errors.
Also,
noting
that
the
Virginia
state
bar
had
opened
an
investigation
into
Guyer’s
conduct,
and
that
Guyer
had
self-reported
the
matter
to
the
Oregon
bar,
the
judge
directed
that
the
transcript
of
the
hearing
be
sent
to
the
bar
authorities
in
those
two
states.
“I
want
those
bar
agencies
to
know
the
ultimate
outcome
here
and
how
I
view
this
issue,”
the
judge
said.
“Hopefully,
that
helps
them
figure
out
what
they
need
to
do,
if
anything.”