Text of WA HB 1951 (2024)

AN ACT Relating to promoting ethical artificial intelligence by 1 protecting against algorithmic discrimination; and adding a new 2 chapter to Title 19 RCW. 3 BE IT ENACTED BY THE LEGISLATURE OF THE STATE OF WASHINGTON: 4 NEW SECTION. Sec. 1. The definitions in this section apply 5 throughout this chapter unless the context clearly requires 6 otherwise. 7 (1) "Algorithmic discrimination" means the condition in which an 8 automated decision tool contributes to unjustified differential 9 treatment or impacts disfavoring people on the basis of race, color, 10 national origin, citizen or immigration status, families with 11 children, creed, religious belief or affiliation, sex, marital 12 status, the presence of any sensory, mental, or physical disability, 13 age, honorably discharged veteran or military status, sexual 14 orientation, gender expression or gender identity, or any other 15 protected class under RCW 49.60.010. 16 (2) "Artificial intelligence" means a machine-based system that 17 can, for a given set of human-defined objectives, make predictions, 18 recommendations, or decisions influencing a real or virtual 19 environment. 20 H-2264.1 HOUSE BILL 1951 State of Washington 68th Legislature 2024 Regular Session By Representative Shavers Prefiled 12/14/23. p. 1 HB 1951 (3) "Automated decision tool" means a system or service that uses 1 artificial intelligence and has been specifically developed and 2 marketed to, or specifically modified to, make, or be a controlling 3 factor in making, consequential decisions. 4 (4) "Consequential decision" means a decision or judgment that 5 has a legal, material, or similarly significant effect on a natural 6 person's life relating to the impact of, access to, or the cost, 7 terms, or availability of, any of the following: 8 (a) Employment, workers management, or self-employment including, 9 but not limited to: 10 (i) Pay or promotion; 11 (ii) Hiring or termination; and 12 (iii) Automated task allocation that automatically limits, 13 segregates, or classifies employees based on individual behavior or 14 performance for the purpose of assigning or determining material 15 terms or conditions of employment; 16 (b) Education and vocational training including, but not limited 17 to: 18 (i) Assessment including, but not limited to, detecting student 19 cheating or plagiarism; 20 (ii) Accreditation; 21 (iii) Certification; 22 (iv) Admissions; and 23 (v) Financial aid or scholarships; 24 (c) Housing or lodging, including rental or short-term housing or 25 lodging; 26 (d) Essential utilities, including electricity, heat, water, 27 internet or telecommunications access, or transportation; 28 (e) Family planning, including adoption services or reproductive 29 services, as well as assessments related to child protective 30 services; 31 (f) Health care or health insurance, including mental health 32 care, dental, or vision; 33 (g) Financial services, including a financial service provided by 34 a mortgage company, mortgage broker, or creditor; 35 (h) The criminal justice system including, but not limited to, 36 risk assessments for pretrial hearings, sentencing, and parole; 37 (i) Legal services, including private arbitration or mediation; 38 (j) Voting; and 39 (k) Access to benefits or services or assignment of penalties. 40 p. 2 HB 1951 (5) "Deployer" means a natural person, partnership, state or 1 local government agency, or corporation that uses or modifies an 2 automated decision tool to make a consequential decision. 3 (6) "Developer" means a natural person, partnership, state or 4 local government agency, or corporation that designs, codes, or 5 produces an automated decision tool, or substantially modifies an 6 artificial intelligence system or service for the known intended 7 purpose of making, or being a controlling factor in making, 8 consequential decisions, whether for its own use or for use by the 9 deployer. 10 (7) "Ethical artificial intelligence" means automated decision 11 tools that are developed and deployed with reasonable efforts by the 12 developer and the deployer to: 13 (a) Minimize unlawful discriminatory or biased outputs or 14 applications; 15 (b) Ensure that automated decision tools are being operated 16 reliably, safely, and consistently; 17 (c) Protect the data of natural persons by incorporating robust 18 privacy and data security measures; 19 (d) Prioritize transparency so that the behavior and functional 20 components of automated decision tools can be understood in order to 21 enable the identification of performance issues, safety and privacy 22 concerns, biases, exclusionary practices, and unintended outcomes; 23 and 24 (e) Promote individual rights and minimize reasonably foreseeable 25 harm to natural persons resulting from use of the automated decision 26 tool. 27 (8) "Impact assessment" means a documented risk-based evaluation 28 of an automated decision tool that meets the criteria of this 29 chapter. 30 (9) "Sex" includes pregnancy, childbirth, and related conditions, 31 gender identity, intersex status, and sexual orientation. 32 (10) "Significant update" means a new version, new release, or 33 other update to an automated decision tool that materially changes 34 its principal use, principal intended use, or expected outcome. 35 NEW SECTION. Sec. 2. (1) By January 1, 2025, and annually 36 thereafter, a deployer of an automated decision tool must complete 37 and document an impact assessment for any automated decision tool the 38 deployer uses that includes all of the following: 39 p. 3 HB 1951 (a) A statement of the purpose of the automated decision tool and 1 its intended benefits, uses, and deployment contexts; 2 (b) A description of the automated decision tool's outputs and 3 how they are used to make, or be a controlling factor in making, a 4 consequential decision; 5 (c) A summary of the types of data collected from natural persons 6 and processed by the automated decision tool when it is used to make, 7 or be a controlling factor in making, a consequential decision; 8 (d) A statement of the extent to which the deployer's use of the 9 automated decision tool is consistent with or varies from the 10 statement required of the developer by section 3 of this act; 11 (e) An assessment of the reasonably foreseeable risks of 12 algorithmic discrimination arising from the use of the automated 13 decision tool known to the deployer at the time of the impact 14 assessment; 15 (f) A description of the safeguards implemented, or that will be 16 implemented, by the deployer to align use of the automated decision 17 tool with principles of ethical artificial intelligence and to 18 address any reasonably foreseeable risks of algorithmic 19 discrimination arising from the use of the automated decision tool; 20 (g) A description of how the automated decision tool will be used 21 by a natural person, or monitored when it is used, to make, or be a 22 controlling factor in making, a consequential decision; and 23 (h) A description of how the automated decision tool has been or 24 will be evaluated for validity or relevance. 25 (2) By January 1, 2025, and annually thereafter, a developer of 26 an automated decision tool must complete and document an impact 27 assessment of any automated decision tool that it designs, codes, or 28 produces that includes all of the following: 29 (a) A statement of the purpose of the automated decision tool and 30 its intended benefits, uses, and deployment contexts; 31 (b) A description of the automated decision tool's outputs and 32 how they are used, as intended, to make, or be a controlling factor 33 in making, a consequential decision; 34 (c) A summary of the types of data collected from natural persons 35 and processed by the automated decision tool when it is used to make, 36 or be a controlling factor in making, a consequential decision; 37 (d) An assessment of the reasonably foreseeable risks of 38 algorithmic discrimination arising from the intended use or 39 foreseeable misuse of the automated decision tool; 40 p. 4 HB 1951 (e) A description of the measures taken by the developer to 1 incorporate principles of ethical artificial intelligence and to 2 mitigate the risk known to the developer of algorithmic 3 discrimination arising from the use of the automated decision tool; 4 and 5 (f) A description of how the automated decision tool is intended 6 to be used by a natural person, or monitored when it is used, to 7 make, or be a controlling factor in making, a consequential decision. 8 (3) A deployer or developer must, in addition to the impact 9 assessment required by subsections (1) and (2) of this section, 10 perform, as soon as feasible, an impact assessment with respect to 11 any significant update. 12 (4) Upon the request of the office of the attorney general, a 13 developer or deployer must provide any impact assessment that it 14 performed pursuant to this section to the office of the attorney 15 general. 16 (5) Impact assessments provided pursuant to subsection (4) of 17 this section are confidential and exempt from disclosure under 18 chapter 42.56 RCW. 19 (6) This section does not apply to a deployer with fewer than 50 20 employees. 21 NEW SECTION. Sec. 3. A developer must provide a deployer with a 22 statement regarding the intended uses of the automated decision tool 23 and documentation regarding all of the following: 24 (1) The known limitations of the automated decision tool, 25 including any reasonably foreseeable risks of algorithmic 26 discrimination arising from its intended use; 27 (2) A description of the types of data used to program or train 28 the automated decision tool; and 29 (3) A description of how the automated decision tool was 30 evaluated for validity and the ability to be explained before sale or 31 licensing. 32 NEW SECTION. Sec. 4. A developer must make publicly available, 33 in a readily accessible manner, a clear policy that provides a 34 summary of both of the following: 35 (1) The types of automated decision tools currently made 36 available to others by the developer; and 37 p. 5 HB 1951 (2) How the developer manages the reasonably foreseeable risks of 1 algorithmic discrimination that may arise from the use of the 2 automated decision tools it currently makes available to others. 3 NEW SECTION. Sec. 5. (1)(a) The attorney general may bring an 4 action in the name of the state, or as parens patriae on behalf of 5 persons residing in the state, to enforce this chapter. For actions 6 brought by the attorney general to enforce this chapter, a violation 7 of this chapter is an unfair or deceptive act in trade or commerce 8 for the purpose of applying the consumer protection act, chapter 9 19.86 RCW. An action to enforce this chapter may not be brought under 10 RCW 19.86.090. 11 (b) The office of the attorney general, before commencing an 12 action under the consumer protection act, chapter 19.86 RCW, must 13 provide 45 days' written notice to a deployer or developer of the 14 alleged violation of this chapter and provide the deployer or 15 developer with an opportunity to cure the alleged violation. The 16 developer or deployer may cure the noticed violation within 45 days 17 of receiving the written notice. 18 (2) Nothing in this chapter shall be construed to limit or 19 otherwise affect the obligations of developers and deployers under 20 applicable laws, rules, or regulations relating to data privacy or 21 security. 22 NEW SECTION. Sec. 6. (1) A deployer may not use an automated 23 decision tool that results in algorithmic discrimination. 24 (2) A violation of this section constitutes an unfair practice 25 under chapter 49.60 RCW, the law against discrimination. All rights 26 and remedies under chapter 49.60 RCW, including the right to file a 27 complaint with the human rights commission and to bring a civil 28 action, apply. 29 NEW SECTION. Sec. 7. Sections 1 through 6 of this act 30 constitute a new chapter in Title 19 RCW. 31 --- END --- p. 6 HB 1951