IOGAN Act

IIB 116TH CONGRESS 1ST SESSION H. R. 4355 IN THE SENATE OF THE UNITED STATES DECEMBER 10, 2019 Received; read twice and referred to the Committee on Commerce, Science, and Transportation AN ACT To direct the Director of the National Science Foundation to support research on the outputs that may be gen- erated by generative adversarial networks, otherwise known as deepfakes, and other comparable techniques that may be developed in the future, and for other pur- poses. VerDate Sep 11 2014 23:40 Dec 10, 2019 Jkt 099200 PO 00000 Frm 00001 Fmt 6652 Sfmt 6652 E:\BILLS\H4355.RFS H4355 kjohnson on DSK79L0C42 with BILLS 2 HR 4355 RFS Be it enacted by the Senate and House of Representa- 1 tives of the United States of America in Congress assembled, 2 SECTION 1. SHORT TITLE. 3 This Act may be cited as the ‘‘Identifying Outputs 4 of Generative Adversarial Networks Act’’ or the ‘‘IOGAN 5 Act’’. 6 SEC. 2. FINDINGS. 7 Congress finds the following: 8 (1) Research gaps currently exist on the under- 9 lying technology needed to develop tools to identify 10 authentic videos, voice reproduction, or photos from 11 manipulated or synthesized content, including those 12 generated by generative adversarial networks. 13 (2) The National Science Foundation’s focus to 14 support research in artificial intelligence through 15 computer and information science and engineering, 16 cognitive science and psychology, economics and 17 game theory, control theory, linguistics, mathe- 18 matics, and philosophy, is building a better under- 19 standing of how new technologies are shaping the 20 society and economy of the United States. 21 (3) The National Science Foundation has iden- 22 tified the ‘‘10 Big Ideas for NSF Future Invest- 23 ment’’ including ‘‘Harnessing the Data Revolution’’ 24 and the ‘‘Future of Work at the Human-Technology 25 VerDate Sep 11 2014 23:40 Dec 10, 2019 Jkt 099200 PO 00000 Frm 00002 Fmt 6652 Sfmt 6201 E:\BILLS\H4355.RFS H4355 kjohnson on DSK79L0C42 with BILLS 3 HR 4355 RFS Frontier’’, in with artificial intelligence is a critical 1 component. 2 (4) The outputs generated by generative adver- 3 sarial networks should be included under the um- 4 brella of research described in paragraph (3) given 5 the grave national security and societal impact po- 6 tential of such networks. 7 (5) Generative adversarial networks are not 8 likely to be utilized as the sole technique of artificial 9 intelligence or machine learning capable of creating 10 credible deepfakes and other comparable techniques 11 may be developed in the future to produce similar 12 outputs. 13 SEC. 3. NSF SUPPORT OF RESEARCH ON MANIPULATED OR 14 SYNTHESIZED CONTENT AND INFORMATION 15 SECURITY. 16 The Director of the National Science Foundation, in 17 consultation with other relevant Federal agencies, shall 18 support merit-reviewed and competitively awarded re- 19 search on manipulated or synthesized content and infor- 20 mation authenticity, which may include— 21 (1) fundamental research on digital forensic 22 tools or other technologies for verifying the authen- 23 ticity of information and detection of manipulated or 24 VerDate Sep 11 2014 23:40 Dec 10, 2019 Jkt 099200 PO 00000 Frm 00003 Fmt 6652 Sfmt 6201 E:\BILLS\H4355.RFS H4355 kjohnson on DSK79L0C42 with BILLS 4 HR 4355 RFS synthesized content, including content generated by 1 generative adversarial networks; 2 (2) fundamental research on technical tools for 3 identifying manipulated or synthesized content, such 4 as watermarking systems for generated media; 5 (3) social and behavioral research related to 6 manipulated or synthesized content, including the 7 ethics of the technology and human engagement 8 with the content; 9 (4) research on public understanding and 10 awareness of manipulated and synthesized content, 11 including research on best practices for educating 12 the public to discern authenticity of digital content; 13 and 14 (5) research awards coordinated with other fed- 15 eral agencies and programs including the Net- 16 working and Information Technology Research and 17 Development Program, the Defense Advanced Re- 18 search Projects Agency and the Intelligence Ad- 19 vanced Research Projects Agency. 20 SEC. 4. NIST SUPPORT FOR RESEARCH AND STANDARDS ON 21 GENERATIVE ADVERSARIAL NETWORKS. 22 (a) IN GENERAL.—The Director of the National In- 23 stitute of Standards and Technology shall support re- 24 search for the development of measurements and stand- 25 VerDate Sep 11 2014 23:40 Dec 10, 2019 Jkt 099200 PO 00000 Frm 00004 Fmt 6652 Sfmt 6201 E:\BILLS\H4355.RFS H4355 kjohnson on DSK79L0C42 with BILLS 5 HR 4355 RFS ards necessary to accelerate the development of the tech- 1 nological tools to examine the function and outputs of gen- 2 erative adversarial networks or other technologies that 3 synthesize or manipulate content. 4 (b) OUTREACH.—The Director of the National Insti- 5 tute of Standards and Technology shall conduct out- 6 reach— 7 (1) to receive input from private, public, and 8 academic stakeholders on fundamental measure- 9 ments and standards research necessary to examine 10 the function and outputs of generative adversarial 11 networks; and 12 (2) to consider the feasibility of an ongoing 13 public and private sector engagement to develop vol- 14 untary standards for the function and outputs of 15 generative adversarial networks or other technologies 16 that synthesize or manipulate content. 17 SEC. 5. REPORT ON FEASIBILITY OF PUBLIC-PRIVATE 18 PARTNERSHIP TO DETECT MANIPULATED OR 19 SYNTHESIZED CONTENT. 20 Not later than 1 year after the date of the enactment 21 of this Act, the Director of the National Science Founda- 22 tion and the Director of the National Institute of Stand- 23 ards and Technology shall jointly submit to the Committee 24 on Space, Science, and Technology of the House of Rep- 25 VerDate Sep 11 2014 23:40 Dec 10, 2019 Jkt 099200 PO 00000 Frm 00005 Fmt 6652 Sfmt 6201 E:\BILLS\H4355.RFS H4355 kjohnson on DSK79L0C42 with BILLS 6 HR 4355 RFS resentatives and the Committee on Commerce, Science, 1 and Transportation a report containing— 2 (1) the Directors’ findings with respect to the 3 feasibility for research opportunities with the private 4 sector, including digital media companies to detect 5 the function and outputs of generative adversarial 6 networks or other technologies that synthesize or 7 manipulate content; and 8 (2) any policy recommendations of the Direc- 9 tors that could facilitate and improve communication 10 and coordination between the private sector, the Na- 11 tional Science Foundation, and relevant Federal 12 agencies through the implementation of innovative 13 approaches to detect digital content produced by 14 generative adversarial networks or other technologies 15 that synthesize or manipulate content. 16 SEC. 6. GENERATIVE ADVERSARIAL NETWORK DEFINED. 17 In this Act, the term ‘‘generative adversarial net- 18 work’’ means, with respect to artificial intelligence, the 19 machine learning process of attempting to cause a gener- 20 ator artificial neural network (referred to in this para- 21 graph as the ‘‘generator’’ and a discriminator artificial 22 neural network (referred to in this paragraph as a ‘‘dis- 23 criminator’’) to compete against each other to become 24 more accurate in their function and outputs, through 25 VerDate Sep 11 2014 23:40 Dec 10, 2019 Jkt 099200 PO 00000 Frm 00006 Fmt 6652 Sfmt 6201 E:\BILLS\H4355.RFS H4355 kjohnson on DSK79L0C42 with BILLS 7 HR 4355 RFS which the generator and discriminator create a feedback 1 loop, causing the generator to produce increasingly higher- 2 quality artificial outputs and the discriminator to increas- 3 ingly improve in detecting such artificial outputs. 4 Passed the House of Representatives December 9, 2019. Attest: CHERYL L. JOHNSON, Clerk. VerDate Sep 11 2014 23:40 Dec 10, 2019 Jkt 099200 PO 00000 Frm 00007 Fmt 6652 Sfmt 6201 E:\BILLS\H4355.RFS H4355 kjohnson on DSK79L0C42 with BILLS