,,ii  #aai? ,,congress  #b;d ,session   
,s4 #dgjg  ,,9 ,,! ,,s5ate ,,( ,,!      
,,unit$ ,,/ates  ,june #h1 #bjbf,mr4    
,coons 7= himself & ,mr4 ,re$7 introduc$
 ! foll[+ bill2 :i* 0 read twice &      
ref]r$ to ! ,committee on ,arm$ ,s]vices
  ,a ,,bill  ,to am5d title #aj1 ,unit$ 
,/ates ,code1 to e/abli% policy = !     
,dep>tm5t ( ,def5se on maximiz+ autonomy
 & artificial intellig5ce sy/ems1 to    
e/abli% requirem5ts relat+ to ,dep>tm5t 
review & v]ification ( autonom\s weapon 
sy/ems & artificial intellig5ce         
capabilities1 & = o!r purposes4  #a4    
,%ort title ,?is ,act may be cit$ z !   
,responsible ,artificial ,intellig5ce   
,def5se ,act ( #bjbf4  #b4 ,policy &    
guidance relat$ to autonom\s weapon     
sy/ems & artificial intellig5ce         
capabilities acquisition  7a7 ,9 g5]al  
,*apt] #cde ( title #aj1 ,unit$ ,/ates  
,code1 is am5d$ by add+ at ! end !      
foll[+ new section3  #degg4 ,autonom\s  
weapon sy/ems & artificial intellig5ce  
capabilities acquisition2 plann+ &      

ov]sight processes  7a7 ,policy ,x is ! 
policy ( ! ,dep>tm5t ( ,def5se to       
maximize uses ( autonomy & artificial   
intellig5ce capabilities to ! ext5t     
practicable1 :ile ensur+ & ma9ta9+ t    
implem5tation ( su* autonomy &          
artificial intellig5ce capabilities     
provides cont9u\s & rigor\s human       
ov]sight to ensure t op]ations are      
conduct$ 9 accordance ) ! law ( w>1     
applicable treaties1 weapon sy/em safety
 rules1 applicable rules ( engagem5t1 & 
long-/&+ frameworks protect+ ! privacy &
 civil lib]ties ( ,unit$ ,/ates p]sons4 
7b7 ,requirem5t ,9 accordance ) ! policy
 set =? 9 subsection 7a71 ! ,secret>y ( 
,def5se %all-- 7#a7 ensure p]sonnel     
ex]cise appropriate levels ( human      
judgm5t & consi/5tly monitor any deploy$
 artificial intellig5ce-enabl$ autonom\s
 weapons sy/ems1 :ile rema9+ responsible
 = ! developm5t1 deploym5t1 & use (     
autonom\s weapon sy/ems & artificial    
intellig5ce capabilities2  7#b7 take    
delib]ate /eps to ensure accuracy 9     

autonom\s weapon sy/ems & artificial    
intellig5ce capabilities2  7#c7 develop 
& deploy autonom\s weapon sy/ems &      
artificial intellig5ce capabilities 9 a 
mann] t promotes an appropriate und]/&+ 
( ! te*nology1 _! developm5t processes1 
& op]ational me?ods applicable to       
autonom\s weapon sy/ems & artificial    
intellig5ce capabilities2  7#d7 subject 
prototype & deploy$ autonomy &          
artificial intellig5ce capabilities to  
r\t9e te/+ & in=mation assurance across 
_! entire life cycles to ensure t       
capabilities meet def9$ safety1         
security1 & effectiv5ess p>amet]s2 &    
7#e7 design & eng9e] autonomy &         
artificial intellig5ce capabilities to  
fulfill _! int5d$ functions1 & deploy ^!
 capabilities su* t human op]ators--    
7,a7 reta9 ! ability to detect & avoid  
un9t5d$ consequ5ces or behaviors2 &     
7,b7 reta9 a means = human int]v5tion to
 dis5gage or deactivate deploy$ sy/ems t
 demon/rate un9t5d$ or illegal behavior4
  7c7 ,review = autonom\s weapons sy/ems

 & support+ artificial intellig5ce      
capabilities  7#a7 ,levels ( human      
judgm5t  7,a7 ,! ,secret>y %all ensure t
 any autonom\s weapon sy/em or          
artificial intellig5ce capability to be 
utiliz$ by ! ,dep>tm5t1 except z exempt$
 by subsection 7c71 is categoriz$ by !  
appropriate level ( human judgm5t       
requir$ to mitigate risks to life1      
safety & heal? ( ,dep>tm5t p]sonnel or  
noncombatant civilian h>m4  7,b7 ,=     
purposes ( subp>agraph 7,a71 appropriate
 levels ( human judgm5t = an autonom\s  
weapon sy/em %all consi/ ( two levels z 
foll[s3 7i7 ,level #a1 :i* means t !    
weapon sy/em poses little to no risk to 
human life or safety %\ld ! sy/em fail  
to act z design$4  7ii7 ,level #b1 :i*  
means t ! weapon sy/em poses mod]ate to 
high risk to human life or safety %\ld !
 sy/em fail to act z design$4  7,c7     
,appropriate levels ( human judgm5t = an
 artificial intellig5ce capability %all 
consi/ ( two levels z foll[s3 7i7 ,level
 #a1 :i* means t ! artificial           

intellig5ce capability poses little to  
no risk z a result ( compromise ( data  
integrity or op]ational support %\ld !  
sy/em fail to act z design$4  7ii7      
,level #b1 :i* means t ! artificial     
intellig5ce capability-- 7,i7 uses data 
protect$ by section #eeb;a ( title #e   
7commonly kn[n z ! ,privacy ,act (      
#aigd7 or regulations promulgat$ "u !   
,heal? ,insurance ,portability &        
,acc\ntability ,act ( #aiif 7,public    
,law #ajd-aia72 or  7,,ii7 poses a risk 
su* t ! failure ( ! artificial          
intellig5ce sy/em to act z design$ w\ld 
sev]ely affect ! ability ( ! ,dep>tm5t  
to p]=m ! designat$ mission ( !         
artificial intellig5ce capability4  7#b7
 ,review & v]ification ,subject to      
subsection 7d71 = any autonom\s weapon  
sy/em or artificial intellig5ce         
capability "u developm5t or be+ field$  
by ! ,dep>tm5t1 ! ,secret>y %all ensure 
t su* autonom\s weapon sy/em or         
artificial intellig5ce capability       
requires-- 7,a7 be=e a decision to ent] 

prototyp+ or =mal developm5t1 review &  
v]ification by ! ,"u ,secret>y ( ,def5se
 = ,rese>* & ,eng9e]+ t-- 7i7 ! sy/em   
design incorporates ! necess>y          
capabilities to all[ comm&]s1 op]ators1 
& analy/s to ex]cise appropriate levels 
( human judgm5t ov] ! use ( =ce 9 !     
envision$ plann+ & employm5t processes =
 ! autonom\s weapon sy/em or artificial 
intellig5ce capability2  7ii7 7,i7 !    
autonom\s weapon sy/em or artificial    
intellig5ce capability is design$ to    
complete engagem5ts )9 a timeframe &    
geographic area1 z well z o!r applicable
 environm5tal & op]ational p>amet]s1    
consi/5t ) comm&] & op]ator int5tions2  
or  7,,ii7 if n design$ z describ$ 9    
subclause 7,i71 ! autonom\s weapon sy/em
 or artificial intellig5ce capability w 
t]m9ate engagem5ts or obta9 additional  
op]ator input be=e cont9u+ ! engagem5t2 
7iii7 ! comb9ation ( ! design & concept 
( employm5t ( ! autonom\s weapon sy/em  
or artificial intellig5ce capability1   
su* z its t>get selection & engagem5t   

logic & o!r relevant processes or       
measures1 acc\nts = risks to nont>gets1 
consi/5t ) comm&] & op]ator int5t & !   
laws ( w>2  7iv7 ! design ( ! autonom\s 
weapon sy/em or artificial intellig5ce  
capability1 includ+ sy/em safety1       
anti-tamp] me*anisms1 & ! cyb]security (
 ! autonom\s weapon sy/em or artificial 
intellig5ce capability1 9 accordance )  
,dep>tm5t ( ,def5se ,in/ruction #hejj4ja
 7relat+ to cyb]security71 or successor 
in/ruction1 addresses & m9imizes !      
probability & consequ5ces ( failures2   
7v7 plans are 9 place = v]ification &   
validation & te/ & evaluation to e/abli%
 ! reliability1 effectiv5ess1 &         
suitability ( ! autonom\s weapon sy/em  
or artificial intellig5ce capability "u 
reali/ic conditions1 includ+ possible   
adv]s>y actions1 to a suffici5t /&>d    
consi/5t ) ! pot5tial consequ5ces ( an  
un9t5d$ engagem5t or unau?oriz$ p>ties  
int]f]+ ) ! op]ation ( ! autonom\s      
weapon sy/em or artificial intellig5ce  
capability prior to field+2 &  7vi7 a   

prelim9>y legal review ( ! autonom\s    
weapon sy/em or artificial intellig5ce  
capability has be5 complet$-- 7,i7 9    
coord9ation ) ! ,g5]al ,c\nsel ( !      
,dep>tm5t ( ,def5se2 &  7,,ii7 9        
accordance ) ,dep>tm5t ( ,def5se        
,directive #ejjj4ja 7relat+ to ,def5se  
,acquisition ,sy/em71 or successor      
directive1 ,dep>tm5t ( ,def5se          
,directive #bcaa4ja 7relat+ to ,dep>tm5t
 ( ,def5se ,law ( ,w> ,program71 or     
successor directive1 &1 :]e applicable1 
,dep>tm5t ( ,def5se ,directive          
#cjjj4jc;,e 7relat+ to ,dep>tm5t (      
,def5se ,ag5t = ,non-,le?al ,weapons &  
,non-,le?al ,weapon ,policy71 or        
successor directive2 &  7,b7 be=e       
field+1 review & v]ification by ! ,"u   
,secret>y = ,rese>* & ,eng9e]+1 9       
consultation ) ! ,vice ,*airman ( !     
,jo9t ,*iefs ( ,/aff1 t1 ) respect to ! 
autonom\s weapon sy/em or artificial    
intellig5ce capability-- 7i7 sy/em      
capabilities1 human-ma*9e int]faces1    
doctr9e1 tactics1 te*niques & proc$ures1

 & tra9+ h be5 demon/rat$ to all[       
comm&]s & op]ators to ex]cise           
appropriate levels ( human judgm5t ov] !
 use ( =ce & to employ sy/ems )         
appropriate c>e & 9 accordance ) ! law (
 w>1 applicable treaties1 weapon sy/em  
safety rules1 & def9able rules (        
engagem5t t are applicable or reasonably
 expect$ to be applicable2  7ii7 sy/em  
safety1 anti-tamp] me*anisms1 cyb]      
survivability1 op]ational resili5ce1 &  
cyb]security capabilities h be5         
implem5t$ to m9imize ! probability &    
consequ5ces ( failures2  7iii7 =        
autonom\s weapon sy/ems or artificial   
intellig5ce capabilities t are be+      
field$ t may h gone ?r\gh ! developm5t  
pipel9e & v]ifi$ "u subp>agraph 7,a71 an
 updat$ legal review ( ! weapon sy/em or
 artificial intellig5ce capability has  
be5 complet$-- 7,i7 9 coord9ation ) !   
,g5]al ,c\nsel ( ! ,dep>tm5t ( ,def5se2 
&  7,,ii7 9 accordance ) ,dep>tm5t (    
,def5se ,directive #ejjj4ja 7relat+ to  
,def5se ,acquisition ,sy/em71 or        

successor directive1 ,dep>tm5t ( ,def5se
 ,directive #bcaa4ja 7relat+ to         
,dep>tm5t ( ,def5se ,law ( ,w>          
,program71 or successor directive1 &1   
:]e applicable1 ,dep>tm5t ( ,def5se     
,directive #cjjj4jc;,e 7relat+ to       
,dep>tm5t ( ,def5se ,ag5t = ,non-,le?al 
,weapons & ,non-,le?al ,weapon ,policy71
 or successor directive2 &  7iv7 a      
monitor+ regime is 9 place to id5tify & 
address *anges 9 op]ational environm5t1 
data inputs1 & use t c\ld contribute to 
failure ( ! sy/em or capability to act 9
 a mann] consi/5t ) ! int5t = ! sy/em or
 capability4  7#c7 ,validity (          
v]ification  7,a7 ,! ,secret>y %all     
treat ea* v]ification "u p>agraph 7#b7  
or p>agraph 7#d7 z valid = a p]iod (    
?ree ye>s4  7,b7 ,an autonom\s weapon   
sy/em or artificial intellig5ce         
capability t is a sub/antially simil>   
v>iant ( ano!r autonom\s weapon sy/em or
 artificial intellig5ce capability t is 
v]ifi$ "u p>agraph 7#b7 or p>agraph 7#d7
 %all also be treat$ z v]ifi$4  7#d7    

,subsequ5t review & v]ification  7,a7 ,=
 any autonom\s weapon sy/em or          
artificial intellig5ce capability t 0   
previ\sly v]ifi$ "u p>agraph 7#b77;,a7  
or exempt$ "u any pr$ecessor review     
process "u ,dep>tm5t ( ,def5se          
,directive #cjjj4ji 7relat+ to ,autonomy
 9 ,weapon ,sy/ems71 t does n curr5tly h
 a valid v]ification pursuant to        
p>agraph 7#c71 ! ,secret>y %all ensure t
 x und]goes subsequ5t review &          
v]ification "u su* p>agraph4  7,b7 ,=   
any autonom\s weapon sy/em or artificial
 intellig5ce capability t 0 previ\sly   
v]ifi$ "u p>agraph 7#b77;,b7 or exempt$ 
"u any pr$ecessor review process "u     
,dep>tm5t ( ,def5se ,directive #cjjj4ji 
7relat+ to ,autonomy 9 ,weapon ,sy/ems7 
t does n curr5tly h a valid v]ification 
pursuant to p>agraph 7#c71 ! ,secret>y  
%all ensure t x und]goes subsequ5t      
review & v]ification "u su* p>agraph z  
if x _h n be5 deploy$4  7#e7 ,privacy   
impact assessm5ts  7,a7 ,= ea*          
artificial intellig5ce capability       

classifi$ "u subsection                 
7c77#a77;,c77ii71 ! ,secret>y ensure t a
 privacy impact assessm5t is conduct$ by
 ! ,director = ,privacy1 ,civil         
,lib]ties & ,transp>5cy4  7,b7 ,9 c>ry+ 
out a privacy impact assessm5t "u       
subp>agraph 7,a71 ! ,director may       
consult ) su* te*nical & policy exp]ts 9
 ! ,dep>tm5t ( ,def5se or else:]e 9 !   
,f$]al ,gov]nm5t z ! ,director consid]s 
appropriate4  7#f7 ,waiv] ( updat$ legal
 review ,! ,"u ,secret>y ( ,def5se =    
,rese>* & ,eng9e]+ may tempor>ily waive 
! requirem5t = an updat$ legal review "u
 p>agraph 7#b77;,b77iii7 = long] ?an "o 
ye> if ! capability is be+ deploy$ 9    
response to real-_w conflict or =       
compell+ national int]e/4  7d7          
,exceptions ,! foll[+ categories (      
autonom\s weapon sy/ems are n subject to
 ! policy set =? 9 subsection 7a7 or !  
requirem5ts ( subsection 7b7 & 7c73 7#a7
 ,op]ator-sup]vis$ autonom\s weapon     
sy/ems us$ to select & engage mat]iel   
t>gets to int]cept attempt$ "t-critical 

or saturation attacks4  7#b7            
,op]ator-sup]vis$ autonom\s weapon      
sy/ems us$ to select & engage mat]iel   
t>gets = def5d+ op]ationally deploy$    
remotely pilot$ or autonom\s vehicles or
 vessels4  7#c7 ,autonom\s or           
semi-autonom\s cyb]space capabilities1  
reasonably judg$ to be non-le?al 9      
nature4  7#d7 ,un>m$ plat=ms1 :e!r      
remotely op]at$ or op]at$ by onbo>d     
p]sonnel1 & :e!r autonom\s or           
semi-autonom\s4  7#e7 ,unguid$          
munitions4  7#f7 ,munitions manually    
guid$ by ! op]ator4  7#g7 ,m9es4  7#h7  
,unexplod$ explosive ordnance4  7#i7    
,autonom\s or semi-autonom\s sy/ems t   
are n weapon sy/ems4  7#aj7 ,any weapon 
sy/em-- 7,a7 t is n cont9u\sly monitor$ 
by a human op]ator2  7,b7 t ! ,secret>y1
 aft] complet+ review ( ! sy/em p]      
subsection 7c71 has det]m9$ to be saf] &
 m reliable = ! int5d$ use ?an          
alt]native sy/ems t are v]ifi$ = su* use
 p] subsection 7c7 & involve cont9u\s   
human sup]vision2 &  7,c7 = :i*-- 7i7 ! 

,secret>y has notifi$ ! congressional   
def5se committees ( ! int5d$ use ( !    
sy/em2 &  7ii7 no congressional def5se  
committee has object$ to dur+ ! #cj-;day
 p]iod beg9n+ on ! date on :i* ! notice 
0 submitt$ "u clause 7i74  7e7          
,prohibitions on c]ta9 uses ( autonomy  
or artificial intellig5ce capabilities  
,except z may be provid$ 9 ano!r /atute1
 ! ,secret>y may n use autonomy or an   
artificial intellig5ce capability = any 
( ! foll[+ use cases3 7#a7 ,= ! decision
 to initiate ! laun* ( a nucle> weapon4 
7#b7 ,= ! monitor+1 track+1 pr(il+1 or  
t>get+ ( an individual or gr\p (        
individuals reasonably believ$ to be 9 !
 ,unit$ ,/ates1 )\t a w>rant obta9$ bas$
 on probable cause ( a crime ) an       
individualiz$1 articulable legal basis1 
or ! collection1 qu]y+1 or analysis (   
in=mation ab\t ! same n o!rwise         
p]missible "u ! ,con/itution ( ! ,unit$ 
,/ates1 reg>dless ( ! orig9 ( ! data    
us$1 except = activities conduct$ 9     
accordance ) applicable provisions (    

law4  7#c7 ,9 ! employm5t ( le?al =ce by
 autonom\s weapon sy/ems )\t            
incorporation ( appropriate levels (    
human judgm5t4  7f7 ,v]ification1       
validation1 te/+ & evaluation (         
autonom\s weapon sy/ems lev]ag+ autonomy
 or artificial intellig5ce sy/ems ,= ea*
 autonomy & artificial intellig5ce sy/em
 t is cov]$ by ! policy set =? 9        
subsection 7a71 reg>dless ( !           
acquisition pa?way or te/ & evaluation  
ov]sight /atus = an autonom\s weapon    
sy/em or artificial intellig5ce         
capability1 ! ,director = ,op]ational   
,te/ & ,evaluation %all ensure !        
autonom\s weapon sy/em or artificial    
intellig5ce capability functions z      
anticipat$ 9 reali/ic op]ational        
environm5ts aga9/ adaptive adv]s>ies &  
are suffici5tly robu/ to m9imize        
failures1 includ+ by ensur+-- 7#a7 su*  
autonom\s weapon sy/em or artificial    
intellig5ce capability goes ?r\gh       
rigor\s h>dw>e & s(tw>e v]ification &   
validation & reali/ic sy/em developm5tal

 & op]ational te/ & evaluation1 includ+ 
analysis ( unanticipat$ em]g5t behavior 
to assess sy/em p]=mance1 capability1   
reliability1 effectiv5ess1 & suitability
 "u reali/ic conditions1 includ+        
possible adv]s>y actions1 consi/5t ) !  
pot5tial consequ5ces ( un9t5d$ engagem5t
 or unau?oriz$ p>ties int]f]+ ) !       
op]ation ( ! sy/em or capability2  7#b7 
h>dw>e & s(tw>e v]ification & validation
 include it]ative cyb] te/ & evaluation 
9 accordance ) ,dep>tm5t ( ,def5se      
,in/ruction #ejjj4hi 7relat+ to ,te/ &  
,evaluation71 or successor in/ruction1  
to v]ify t autonom\s weapon sy/em or    
artificial intellig5ce capability is    
resili5t & survivable 9 conte/$         
cyb]space1 if relevant to ! purpose &   
mission ( ! sy/em or capability2  7#c7  
sy/ems incorporat+ autonomy or          
artificial intellig5ce capabilities g   
?r\gh rigor\s developm5tal & op]ational 
te/ & evaluation to v]ify & validate t  
autonom\s weapon sy/em or artificial    
intellig5ce capability is robu/ accord+ 

to design requirem5ts2  7#d7 te/ &      
evaluation ( sy/ems incorporat+ autonomy
 or artificial intellig5ce capabilities 
include te/+ to confirm t _! autonomy or
 artificial intellig5ce algori?ms c be  
rapidly reprogramm$ on new input data to
 enable timely correction ( any un9t5d$ 
sy/em behaviors t may be obs]v$ or      
discov]$ dur+ future sy/em op]ations2   
7#e7 adequate tra9+1 tactics1 te*niques1
 proc$ures1 & doctr9e are available1    
p]iodically review$1 & us$ by sy/em     
op]ators & comm&]s to und]/& !          
function+1 capabilities1 & limitations (
 ! sy/em's autonomy or artificial       
intellig5ce 9 reali/ic op]ational       
conditions2  7#f7 sy/em design &        
human-ma*9e int]faces are readily       
und]/&able to tra9$ op]ators1 ) cle>    
ability = tra9$ op]ators to activate &  
deactivate sy/em functions2  7#g7 aft]  
initial op]ational te/+ & evaluation1 z 
direct$ by ! ,director1 sy/em data is   
collect$ & any fur!r *anges to ! sy/em  
und]go appropriate v]ification1         

validation1 te/1 & evaluation to ensure 
t critical safety features h n be5      
degrad$2  7#h7 sy/em s(tw>e is te/$ us+ 
! be/ means & me?ods available to !     
,dep>tm5t to validate t critical safety 
features h n be5 degrad$2  7#i7 automat$
 te/+ tools1 su* z model+ & simulation1 
are us$ :5ev] feasible2  7#aj7 te/+     
id5tifies any new op]at+ /ates & o!r    
relevant *anges 9 ! autonom\s weapon    
sy/em or artificial intellig5ce         
capability2  7#aa7 z direct$ by !       
,director-- 7,a7 ea* new or revis$      
op]at+ /ate und]goes appropriate &      
tailor$ additional te/ & evaluation to  
*>act]ize ! sy/em behavior 9 t new      
op]at+ /ate2 &  7,b7 :ole sy/em foll[-on
 op]ational te/ & evaluation :5 requir$ 
due to *anges to ! /ate transition      
matrix2 &  7#ab7 9 coord9ation ) ! ,"u  
,secret>y = ,rese>* & ,eng9e]+ &        
,director1 ! own+ compon5t ( ! ,dep>tm5t
 provides = monitor+ to id5tify &       
address :5 *anges to ! sy/em design or  
op]ational environm5t require additional

 te/+ & evaluation to provide suffici5t 
confid5ce t ! sy/em w cont9ue to avoid  
un9t5d$ engagem5ts & resi/ int]f]5ce by 
unau?oriz$ p>ties4  7g7 ,autonomy &     
,artificial ,intellig5ce ,sy/ems ,work+ 
,gr\p  7#a7 ,e/abli%m5t  7,a7 ,!        
,secret>y %all e/abli% & *>t] a work+   
gr\p = ! purposes set =? 9 p>agraph     
7#b74  7,b7 ,! work+ gr\p e/abli%$ "u   
subp>agraph 7,a7 %all be kn[n z !       
,autonomy & ,artificial ,intellig5ce    
,work+ ,gr\p4  7#b7 ,purposes ,!        
purposes set =? 9 ?is p>agraph are z    
foll[s3 7,a7 ,to support ! ,"u ,secret>y
 ( ,def5se = ,rese>* & ,eng9e]+1 & !    
,vice ,*airman ( ! ,jo9t ,*iefs ( ,/aff 
9 consid]+ ! full range ( relevant      
,dep>tm5t int]e/s dur+ ! review (       
autonom\s weapon sy/ems & artificial    
intellig5ce capabilities be=e =mal      
developm5t4  7,b7 ,to support ! ,"u     
,secret>y ( ,def5se = ,rese>* & ,eng9e]+
 & ! ,vice ,*airman ( ! ,jo9t ,*iefs (  
,/aff1 9 consid]+ ! full range (        
relevant ,dep>tm5t int]e/s dur+ ! review

 ( autonom\s weapon sy/ems be=e         
deploym5t4  7,c7 ,:5 reque/$ by         
appropriate repres5tatives ( !          
secret>ies ( ! milit>y dep>tm5ts1 !     
,comm&] ( ,unit$ ,/ates ,special        
,op]ations ,comm&1 or1 :5 applicable1 a 
director ( a def5se ag5cy or a ,dep>tm5t
 ( ,def5se ,field ,activity-- 7i7 to    
advise :e!r a giv5 weapon sy/em requires
 s5ior-level approval 9 accordance ) ?is
 section2 &  7ii7 to help id5tify &     
advise on address+ pot5tial issues      
pres5t$ by a giv5 weapon sy/em dur+ a   
pot5tial s5ior-level review 9 accordance
 ) ?is section4  7,d7 ,to develop &     
issue safety /&>ds = use ( autonomy &   
artificial intellig5ce capabilities 9   
evaluation ( su* capabilities4  7h7     
,annual report ,n lat] ?an ,janu>y #ca (
 ea* ye> until ,janu>y #ca1 #bjcg1 !    
,secret>y %all submit to ! congressional
 def5se committees an annual report on !
 adm9i/ration ( ?is section4  7i7       
,def9itions ,9 ?is section3 7#a7 ,! t]m 
artificial intellig5ce has ! mean+ giv5 

! t]m section #ejjb ( ! ,national       
,artificial ,intellig5ce ,initiative    
,act ( #bjbj 7#ae ,u4,s4,c4 #idja74     
7#b7 ,! t]m autonom\s weapon sy/ems     
means a weapon sy/em t1 once activat$1 c
 select & engage t>gets )\t fur!r       
int]v5tion by an op]ator4 ,su* t]m      
includes op]ator-sup]vis$ autonom\s     
weapon sy/ems t are design$ to all[     
op]ators to ov]ride op]ation ( ! weapon 
sy/em b c select & engage t>gets )\t    
fur!r op]ator input aft] activation4    
7#c7 ,! t]m autonomy means a capability 
7or set ( capabilities7 t enables a     
p>ticul> action ( a sy/em to be         
automatic or1 )9 specifi$ b\nd>ies      
self-gov]n+ ) m9imal human ov]sight4  4 
7b7 ,cl]ical am5dm5t ,! table ( sections
 = *apt] #cde ( title #aj1 ,unit$ ,/ates
 ,code is am5d$ by ins]t+ aft] ! item   
relat$ to section #degf ! foll[+ new    
item3  #degg4 ,autonom\s weapon sy/ems &
 artificial intellig5ce capabilities    
acquisition2 plann+ & ov]sight          
processes4  4