AI Frontier Institute

AI-MEANINGFUL HUMAN REVIEW

View primary source →

Illinois HB3720, introduced by Rep. Abdelnasser Rashid in the 104th General Assembly, creates the Meaningful Human Review of Artificial Intelligence Act (Section 1). The bill prohibits State agencies, or entities acting on their behalf, from using any automated decision-making system without continuous meaningful human review when performing functions related to public assistance benefits, functions that materially impact individual rights, civil liberties, safety, or welfare, or functions affecting statutorily or constitutionally provided rights (Section 10(a)). State agencies are also barred from procuring such systems unless they include continuous meaningful human review (Section 10(b)). The bill explicitly protects existing employee rights, including collective bargaining agreements, and prohibits the displacement of employees or transfer of their duties to automated systems (Section 10(c)). State agencies seeking to use a permitted automated decision-making system must conduct an impact assessment signed by those responsible for meaningful human review, repeated at least every two years and before any material system change (Section 15(a)). Impact assessments must include: a description of the system's objectives; an evaluation of its ability to achieve those objectives; a summary of underlying algorithms and training data; testing for accuracy, fairness, bias, discrimination, cybersecurity vulnerabilities, privacy risks, public health and safety risks, foreseeable misuse, and data handling; and notification mechanisms informing affected individuals of their rights (Section 15(a)(1)-(5)). If an impact assessment finds discriminatory or biased outcomes, the agency must cease all use of the system (Section 15(b)). Each assessment must be submitted to the Governor, President of the Senate, and Speaker of the House at least 30 days before system implementation and published on the agency's website, with limited redaction permitted for security, privacy, or operational reasons (Section 20).

← Back to the tracker