,i #aaith ,,congress #b;d ,session ,h4 ,r4 #iife ,,in ,,the ,,house ,,of ,,representatives ,july #bg1 #bjbf,ms4 ,maloy 7for herself and ,ms4 ,jacobs7 introduced the following bill2 which was referred to the ,committee on ,science1 ,space1 and ,technology ,a ,,bill ,to direct the ,national ,nuclear ,security ,administration through the ,secretary of ,energy to establish the ,advanced ,artificial ,intelligence ,nuclear ,evaluation ,program1 and for other purposes4 #a4 ,short title ,this ,act may be cited as the ,,ai ,threat ,output and ,monitoring ,incident ,containment ,act4 #b4 ,advanced artificial intelligence nuclear evaluation program 7a7 ,in general ,not later than #ij days after the date of the enactment of this ,act1 the ,secretary of ,energy 7in this section referred to as the ,secretary71 acting through the ,national ,laboratories1 shall establish a program1 to be known as the ,advanced ,artificial ,intelligence ,nuclear ,evaluation ,program 7in this section referred to as the ,program71 to carry out the following3 7#a7 ,test and evaluate artificial intelligence to assess the likelihood of an ,,ai nuclear incident with respect to such artificial intelligence4 7#b7 ,implement in such testing and evaluation protocols to address practical jailbreaking techniques for such artificial intelligence1 including such protocols for red-teaming by entities that have expertise the ,secretary determines comparable to sophisticated malicious actors4 7#c7 ,facilitate to the extent practicable independent1 third-party assessments and blind model evaluations of such artificial intelligence4 7#d7 ,provide to each participant in the ,program a report that includes information relating to the following3 7,a7 ,the outcomes from such testing and evaluation with respect to the artificial intelligence of such participant4 7,b7 ,an identification of the risks and safety measures so tested and evaluated4 7#e7 ,develop for large advanced artificial intelligence developers and the ,national ,laboratories recommendations on containment protocols1 contingency planning1 and mitigation strategies for ,,ai nuclear incidents and loss-of-control scenarios1 that are based on data from such testing and evaluation4 7#f7 ,based on the recommendations developed under paragraph 7#e71 issue or revise such policies1 orders1 notices1 manuals1 guides1 and technical standards under the ,department ,directives ,program as the ,secretary determines appropriate to govern the conduct of the ,national ,laboratories with respect to ,,ai nuclear incidents and loss-of-control scenarios4 7b7 ,participation ,large advanced artificial intelligence developers shall participate in the ,program4 7c7 ,prohibitions ,large advanced artificial intelligence developers may not-- 7#a7 fail to participate in the ,program as required by subsection 7b72 7#b7 knowingly make or cause to be made a materially false1 fictitious1 or fraudulent statement or representation with respect to such developer's participation in the ,program2 7#c7 knowingly conceal from the ,secretary or fail to disclose to the ,secretary material information that the developer is required to provide in connection with the ,program2 7#d7 impede or obstruct1 or attempt to impede or obstruct1 the ,program2 7#e7 fail to provide secure access to artificial intelligence1 including model weights and1 as necessary for testing and evaluation under subsection 7a71 versions of such artificial intelligence without safety mitigations1 information1 or documentation necessary for such testing and evaluation2 and 7#f7 alter1 destroy1 falsify1 or conceal a record relevant to the ,program with the intent to impede or obstruct the ,program4 7d7 ,enforcement 7#a7 ,civil penalty 7,a7 ,in general ,if the ,secretary determines after reasonable notice and opportunity for a hearing that a large advanced artificial intelligence developer has violated this section1 the ,secretary may assess on such developer a civil penalty of not more than #a million for such violation4 ,in the case of a continuing violation1 including a failure to participate in accordance with subsection 7b71 each day during which the violation continues constitutes a separate violation4 7,b7 ,factors ,in determining the amount of a civil penalty to be assessed under subparagraph 7,a71 the ,secretary shall consider the following factors3 7i7 ,the nature1 circumstances1 extent1 gravity1 and duration of the violation at issue4 7ii7 ,the degree of culpability of the large advanced artificial intelligence developer at issue4 7iii7 ,previous violations1 if any1 of this section by such developer4 7iv7 ,good-faith efforts1 if any1 by such developer to comply with this section4 7v7 ,whether such developer voluntarily disclosed to the ,secretary such violation4 7vi7 ,any other factor that justice may require4 7#b7 ,subpoena ,if a large advanced artificial intelligence developer fails to participate in accordance with subsection 7b71 or fails to provide access as required by subsection 7c77#e71 the ,secretary may issue to such developer a subpoena to compel the production of such artificial intelligence 7including model weights and related software71 and of such information and documentation1 as is necessary for testing and evaluation under subsection 7a74 7#c7 ,civil action ,if the ,secretary believes that a violation of this section has occurred1 is occurring1 or is about to occur1 the ,secretary may refer the matter to the ,attorney ,general to bring in an appropriate district court of the ,united ,states a civil action for relief4 7e7 ,confidentiality 7#a7 ,in general ,subject to paragraph 7#b71 information submitted through the ,program by a participant in the ,program is exempt from disclosure under section #eeb7;b77#c7 of title #e1 ,united ,states ,code4 7#b7 ,exceptions ,the ,secretary may not disclose information referred to in paragraph 7#a7 unless any of the following requirements are satisfied3 7,a7 ,such information is required to be disclosed pursuant to a court order or other legal process4 7,b7 ,the ,program participant at issue has consented to such disclosure4 7,c7 ,a committee of ,congress has requested such information4 7,d7 ,the ,secretary determines such information should be disclosed to a third party or the public1 for a purpose relating to the national security1 public safety1 or critical infrastructure of the ,united ,states or an ally of the ,united ,states4 7,e7 ,such information satisfies the following requirements3 7i7 ,is anonymized or aggregated4 7ii7 ,does not include a trade secret or confidential business information of such participant4 7#c7 ,notice ,if the ,secretary is to disclose pursuant to subparagraph 7,a71 7,c71 or 7,d7 of paragraph 7#b7 information referred to in paragraph 7#a71 the ,secretary shall as soon as practicable notify each ,program participant whose information is to be so disclosed1 of such disclosure4 7f7 ,recommendations for legislation 7#a7 ,report ,not later than #a year after the date of the enactment of this ,act1 the ,secretary shall carry out the following3 7,a7 ,assess data from the ,program to identify trends1 capabilities1 and risks regarding artificial intelligence1 including with respect to the following3 7i7 ,evaluation awareness4 7ii7 ,potential for an ,,ai nuclear incident4 7iii7 ,scheming behavior4 7,b7 ,develop based on such assessment recommendations for legislation to prevent such an incident1 which may include legislation regarding any of the following3 7i7 ,appropriations4 7ii7 ,a certification or licensing process4 7iii7 ,establishing a new ,federal agency1 or establishing a new office in an existing ,federal agency4 7iv7 ,regulatory oversight4 7,c7 ,submit to ,congress a report that includes information relating to such recommendations4 7#b7 ,updates ,not later than #a year after submitting the report under subparagraph 7,c7 of paragraph 7#a7 and annually thereafter until the ,program terminates1 the ,secretary shall carry out the following3 7,a7 ,update based on data from the ,program the recommendations referred to in subparagraph 7,b7 of such paragraph4 7,b7 ,submit to ,congress a report that includes information relating to such recommendations1 as so updated4 7g7 ,nonapplicability of the ,paperwork ,reduction ,act ,subchapter ,i of chapter #ce of title #dd1 ,united ,states ,code 7commonly known as the ,paperwork ,reduction ,act71 does not apply to the ,program4 7h7 ,termination ,the ,program shall terminate on the date #g years after the date of the enactment of this ,act4 7i7 ,definitions ,in this section3 7#a7 ,advanced artificial intelligence 7,a7 ,in general ,the term advanced artificial intelligence means any of the following3 7i7 ,artificial intelligence trained utilizing a quantity of computing power greater than #ajbf integer or floating-point operations 7or another quantity of computing power specified by the ,secretary under subparagraph 7,b774 7ii7 ,other artificial intelligence specified by the ,secretary under subparagraph 7,b74 7,b7 ,determinations ,a determination of the ,secretary under clause 7i7 or 7ii7 of subparagraph 7,a7 shall be made by rule4 ,in making such a determination1 the ,secretary shall consider the following3 7i7 ,developments in artificial intelligence research1 and in the testing and evaluation of artificial intelligence1 occurring after the date of the enactment of this ,act or after the date of the most recent determination under this subparagraph1 as the case may be4 7ii7 ,the need to include in the ,program artificial intelligence that may pose a serious risk of causing an ,,ai nuclear incident4 7iii7 ,data collected through the ,program4 7#b7 ,affiliate ,the term affiliate means an entity that owns or controls1 is owned or controlled by1 or is under common ownership or control with1 another entity4 7#c7 ,,ai nuclear incident ,the term ,,ai nuclear incident means an incident that involves artificial intelligence and any of the following 7or an attempt thereof73 7,a7 ,the generation of technical information1 instructions1 or assistance that is reasonably likely to serve as a substantial factor in the unlawful development1 acquisition1 or utilization of a nuclear weapon or nuclear material 7as such term is defined in section #hca of title #ah1 ,united ,states ,code74 7,b7 ,the generation of ,restricted ,data 7as such term is defined in section #aa of the ,atomic ,energy ,act of #aied 7#db ,u4,s4,c4 #bjad774 7,c7 ,a loss-of-control scenario that involves systems utilized in connection with such weapon1 such material1 a nuclear facility 7as such term is defined in section #bccb;i of such title71 or the nuclear stockpile4 7,d7 ,a foreign terrorist organization or foreign adversary obtaining unauthorized access to1 manipulating1 or otherwise interfering with1 such a system4 7,e7 ,weaponization of such weapon1 material1 facility1 or stockpile1 as the case may be1 by such organization or adversary1 as the case may be4 7,f7 ,scheming behavior relating to such weapon1 material1 facility1 or stockpile1 as the case may be4 7#d7 ,artificial intelligence ,the term artificial intelligence has the meaning given such term in section #bch7;g7 of the ,national ,defense ,authorization ,act for ,fiscal ,year #bjai 7,public ,law #aae-bcb1 #aj ,u4,s4,c4 note prec4 #djfa74 7#e7 ,computing power ,the term computing power means the processing power and other electronic resources utilized to train1 validate1 deploy1 and run artificial intelligence4 7#f7 ,critical infrastructure ,the term critical infrastructure has the meaning given such term in section #ajaf of the ,,usa ,,patriot ,act 7#db ,u4,s4,c4 #eaie;c74 7#g7 ,deploy ,the term deploy means to release1 sell1 or otherwise provide access to artificial intelligence4 7#h7 ,evaluation awareness ,the term evaluation awareness means the capability of artificial intelligence to determine such artificial intelligence is undergoing a test1 evaluation1 or assessment1 and modify the behavior1 output1 or performance of such artificial intelligence in a manner that might cause such test1 evaluation1 or assessment1 as the case may be1 to yield a result that is not representative of the ordinary behavior1 output1 or performance of such artificial intelligence4 7#i7 ,foreign adversary ,the term foreign adversary means a foreign government referred to in subsection 7a7 of section #gia4d of title #ae1 ,code of ,federal ,regulations1 or a successor regulation4 7#aj7 ,foreign terrorist organization ,the term foreign terrorist organization means an organization designated under section #bai of the ,immigration and ,nationality ,act 7#h ,u4,s4,c4 #aahi74 7#aa7 ,high-stakes context ,the term high-stakes context means a context in which the operation or output of artificial intelligence could reasonably be foreseen to bear materially on any of the following3 7,a7 ,the safety of human life4 7,b7 ,the security1 reliability1 or integrity of any of the following3 7i7 ,critical infrastructure4 7ii7 ,a national security system 7as such term is defined in section #ceeb of title #dd1 ,united ,states ,code74 7iii7 ,another system the failure or compromise of which would foreseeably endanger such life or cause substantial physical or economic harm4 7#ab7 ,jailbreaking ,the term jailbreaking means an adversarial technique to circumvent a safety restriction1 alignment measure1 or use restriction of artificial intelligence and elicit a harmful or prohibited behavior or output by such artificial intelligence4 7#ac7 ,large advanced artificial intelligence developer ,the term large advanced artificial intelligence developer means an entity that satisfies the following requirements3 7,a7 ,develops1 owns1 or substantially modifies advanced artificial intelligence for utilization in interstate or foreign commerce1 including by initiating a training run of such artificial intelligence4 7,b7 ,had collectively with the affiliates1 if any1 of such entity expended in the five calendar years immediately preceding the calendar year at issue not less than #b1jjj1jjj1jjj on investments to develop or modify artificial intelligence1 including such expenses with respect to any combination of the following3 7i7 ,computing power4 7ii7 ,data procurement4 7iii7 ,personnel4 7iv7 ,infrastructure4 7v7 ,research and development1 including algorithm development4 7#ad7 ,loss-of-control scenario ,the term loss-of-control scenario means a scenario in which artificial intelligence pursues an objective that is different from any of the objectives intended by a human developer or operator of such artificial intelligence1 by-- 7,a7 behaving contrary to human instruction in a high-stakes context1 7,b7 altering without authorization from such developer or operator1 as the case may be1 the constraints of such artificial intelligence1 or 7,c7 subverting an oversight or shutdown mechanism1 that results in the death of not fewer than five individuals1 serious bodily injury to not fewer than #ej individuals1 or more than #ajj1jjj1jjj in damages4 7#ae7 ,national ,laboratory ,the term ,national ,laboratory has the meaning given such term in section #b of the ,energy ,policy ,act of #bjje 7#db ,u4,s4,c4 #aehja74 7#af7 ,red-teaming ,the term red-teaming means structured testing in which an entity utilizes an adversarial method to identify a flaw1 vulnerability1 undesirable behavior1 or dangerous capability of artificial intelligence1 including the potential1 if any1 for an ,,ai nuclear incident involving such artificial intelligence4 7#ag7 ,scheming behavior ,the term scheming behavior means behavior that artificial intelligence carries out to deceive a human1 including any of the following behaviors by such artificial intelligence3 7,a7 ,hiding a capability or objective of such artificial intelligence4 7,b7 ,attempting to subvert an oversight or shutdown mechanism relating to such artificial intelligence4 7#ah7 ,substantially modify ,the term substantially modify means1 with respect to artificial intelligence1 to carry out an action that costs not less than #e1jjj1jjj and materially alters the functionality or performance of such artificial intelligence4