Analyzing the Colorado AI Act
David Stauss, Partner, CIPP/US/E, CIPT, FIP, PLS
Erik Dullea, Partner, CIPP/US, CIPMM
Shelby Dolen, Attorney, CIPP/US
Owen Davis, Attorney
David Stauss, Partner, CIPP/US/E, CIPT, FIP, PLS
Erik Dullea, Partner, CIPP/US, CIPMM
Shelby Dolen, Attorney, CIPP/US
Owen Davis, Attorney
1. Background and Timeline
Scope and Relevant Definitions 3. Developer Obligations 4. Deployer Obligations 5. AI Disclosure Obligations
Exemptions
What to Do Now 8. Other AI Laws & Pending Bills
Organized by Connecticut Senator James Maroney
Facilitated by Future of Privacy Forum
Bi-partisan group of lawmakers from nearly 30 states
Met seven times during last summer and fall
Heard from AI experts from across multiple fields and geographies
“In recent years, algorithmic decision-making has produced biased, discriminatory, and otherwise problematic outcomes in some of the most important areas of the American economy. ”
“Mounting evidence reveals that algorithmic decisions can produce biased, discriminatory, and unfair outcomes in a variety of high-stakes economic spheres including employment, credit, health care, and housing.”
FTC Commissioner Rebecca Kelly Slaughter, Algorithms and Economic Justice: A Taxonomy of Harms and a Path Forward for the Federal Trade Commission, Yale Journal of Law & Technology, August 2021
• Regulating in complex and rapidly changing area hurts businesses
• Regulation hurts innovation
• Patchwork of state laws hurts businesses
• US cannot fall behind like it did with privacy regulation
• Guardrails can provide for responsible innovation
• Businesses can be protected through rebuttable presumption, affirmative defense, and no private right of action
• Law is limited in scope (current focus is on discrimination in high-risk applications)
Responsible deployment of high-risk AI systems
Provide disclosures and notices Review as needed
Provide consumer with rights
Notify relevant parties if highrisk AI discriminates
May 17, 2024
• Signed into law
Aug. 1, 2024
• Task force creation deadline
Feb. 1, 2025
• Task force report deadline
1 year, 8 months, 15 days
Feb. 1, 2026
• Effective
Artificial Intelligence System Deployed Makes a consequential decision
Is a substantial factor in making a consequential decision OR
“Any machine-based system that, for any explicit or implicit objective, infers from the inputs the system receives how to generate outputs, including content, decisions, predictions, or recommendations, that can influence physical or virtual environments”
To “use a high-risk artificial intelligence system”
Legal service
Education enrollment or education opportunity
Insurance
Decision that has material legal or similarly significant effect on provision or denial to any consumer of, or cost or terms of:
Employment or employment opportunity
Healthcare services Housing
Financial or lending service
Essential government service
A factor that:
1. assists in making a consequential decision;
2. is capable of altering the outcome of a consequential decision; and
3. is generated by an artificial intelligence system
Any “use of an artificial intelligence system to generate any content, decision, prediction, or recommendation concerning a consumer that is used as a basis to make a consequential decision concerning the consumer”
• perform a narrow procedural task, or
• detect decision-making patterns or deviations from prior decision-making patterns and is not intended to replace or influence a previously completed human assessment without sufficient human review
Unless it makes or is a substantial factor in making a consequential decision:
• Anti-fraud technology that does not use facial recognition technology
• Anti-malware
• Anti-virus
• Artificial intelligence-enabled video games
• Calculators
• Cybersecurity
• Databases
• Data storage
• Firewall
• Internet domain registration
• Internet website loading
• Networking
• Spam and robocall-filtering
• Spell-checking
• Spreadsheets
• Web caching
• Web hosting or any similar technology; or
• Technology that communicates with consumers in natural language for the purpose of providing users with information, making referrals or recommendations, and answering questions and is subject to an accepted use policy that prohibits generating content that is discriminatory or harmful.
• Person doing business in Colorado that develops or intentionally and substantially modifies an AI system
• “Intentional and substantial modification”
• Deliberate change made to AI system that results in new reasonably foreseeable risk of algorithmic discrimination
• Subject to exceptions
• Person doing business in Colorado that uses high-risk AI system
Notify of algorithmic discrimination
Disclosures to deployers
Public disclosures
Standard
• Developer of high-risk AI system shall use reasonable care to protect consumers from any known or reasonably foreseeable risks of algorithmic discrimination arising from the intended and contracted uses of the high-risk AI system.
Rebuttable Presumption
• If developer follows law’s requirements, there is a rebuttable presumption that it used reasonable care.
• Any condition in which the use of an artificial intelligence system results in an unlawful differential treatment or impact that disfavors an individual or group of individuals on the basis of their actual or perceived age, color, disability, ethnicity, genetic information, limited proficiency in the English language, national origin, race, religion, reproductive health, sex, veteran status, or other classification protected under the laws of this state or federal law.
• Self Testing: Use of high-risk AI system for sole purpose of self-testing to identify, mitigate, or prevent discrimination or otherwise ensure compliance with state and federal law;
• Promote Diversity: Use of high-risk AI system for sole purpose of expanding an applicant, customer, or participant pool to increase diversity or redress historical discrimination; or
• Private Club: Act or omission by or on behalf of a private club or other establishment that is not in fact open to the public, as set forth in Title II of the federal Civil Rights Act of 1964
General Statement
• Describing reasonably foreseeable uses and known harmful or inappropriate uses of system
Documentation
Disclosing
• Summaries of data used to train system
• Known or reasonably foreseeable limits of system
• Purpose of system
• Intended benefits and uses
• Information necessary for deployer to comply with its obligations, including information necessary for impact assessments
• How system was evaluated to mitigate algorithmic discrimination
• Data governance measures
• Intended outputs of system
Documentation
Describing
• Measures taken to mitigate known or reasonably foreseeable risks of algorithmic discrimination
• How system should and should not be used, and be monitored
Additional
Documentation
• Any additional documentation reasonably necessary to assist deployer in understanding outputs and to monitor performance of system
• Types of high-risk AI systems developer has deployed or intentionally and substantially modified and makes available to a deployer or other developer
• How developer manages known or reasonably foreseeable risks of algorithmic discrimination that may arise from types of highrisk AI systems
Within 90 days, developer must notify Attorney General and all known deployers or other developers of the high-risk AI system, of any known or reasonably foreseeable risk of algorithmic discrimination arising from intended uses of system that developer learns through (1) its testing or (2) credible reports from a deployer
Standard
• Deployer of high-risk artificial intelligence system shall use reasonable care to protect consumers from any known or reasonably foreseeable risks of algorithmic discrimination.
Rebuttable Presumption
• If developer follows law’s requirements, there is a rebuttable presumption that it used reasonable care.
Risk Identification and Mitigation for Deployment of High-Risk AI Systems
• Specify and incorporate principles, processes and personnel used to identify, document and mitigate known or reasonably foreseeable risks of algorithmic discrimination
Review and Update
• Be an iterative process subject to systematic, regular review and updating during the high-risk AI system’s life cycle
Reasonable
• Be reasonable considering guidance and standards from NIST AI framework or other framework recognized by AG; size and complexity of deployer; nature and scope of highrisk AI system; and sensitivity and data processed by system
Statement of purpose, intended uses cases, deployment context of, and benefits afforded by system
Whether system poses known or reasonably foreseeable risk of algorithmic discrimination and, if so, steps taken to mitigate risks
Description of categories of data system processes as inputs and outputs system produces
If deployer used data to customize system, an overview of categories of data
Any metrics use to evaluate performance and known limitations of system
Description of transparency measures taken, including measures taken to disclose system to consumers
Description of post-deployment monitoring and user safeguards
At least annually, the deployer or a third party contracted by the deployer, must review the deployment of each high-risk AI system to ensure that it is not causing algorithmic discrimination.
• Deployer must notify consumers that it has deployed system to make, or be a substantial factor in making, a consequential decision
• Provide statement disclosing purpose of system, nature of consequential decision, contact information, plain language description of system and how to access public statement (discussed below)
• Providing information regarding right to opt out of profiling under Colorado Privacy Act, if applicable
Degree to which, and manner in which, system contributed to consequential decision
Type of data that was processed by system to make decision
Source of data
• Opportunity to correct any incorrect personal data that system processed to make decision
Appeal
• Opportunity to appeal decision unless not in best interest of consumer
• If technically feasible, allow for human review
• Types of high-risk AI systems that deployer currently deploys
• How deployer manages known or reasonably foreseeable risks of algorithmic discrimination
• In detail, the nature, source, and extent of the information collected and used by the deployer
• Deployer does not have to comply with risk management policy and program, impact assessment, and public disclosure requirements if:
• It has less than 50 full-time equivalent employees and does not use deployer’s own data to train system; and
• System is used for intended uses disclosed by developer, system continues learning based on data derived from sources other than deployer’s own data and deployer makes available any impact assessment that developer provides that includes required information
Within 90 days, deployer must notify Attorney General if deployer discovers high-risk AI system has caused algorithmic discrimination
Deployers and certain developers that make available an AI system intended to interact with consumers shall disclose to consumers that interact with the system that it is an AI system
Exception
It would be obvious to a reasonable person that they are interacting with AI system
Comply with law
Cooperate with law enforcement
Investigate/defend legal claims
Prevent, detect, respond to security incidents, fraud, etc. (except for use of facial recognition technology)
Engage in certain types of research
Effectuate product recall
Identify and repair technical errors that impair existing or intended functionality
High-risk AI system approved, authorized, certified, cleared, developed or granted by federal agency
High-risk AI system in compliance with substantially equivalent standards established by federal agency
Conducting research to support application for approval or certification from federal agency
Performing work under, or in connection with, contract with certain federal departments unless used for employment or housing
AI system acquired by or for federal government or federal agency or department unless used for employment or housing
• Used by covered entity and is providing healthcare recommendations that (1) are generated by AI system, (2) require healthcare provider to take action to implement the recommendations and (3) are not high-risk.
• Insurer, fraternal benefit society or developer of AI system used by insurer if subject to C.R.S. § 10-3-1104.9 and rules adopted by Commissioner of Insurance
• Bank, out-of-state bank, credit union, out-of-state credit union, or any affiliate or subsidiary if subject to regulatory regime that is substantially equivalent to law and includes auditing and risk mitigation
Attorney General Enforcement
• Civil penalty of up to $20,000 per violation
• No private right of action
• No district attorney enforcement
Affirmative Defense – Deployer will have the Burden
• (1) Discover and cure violation and is also
• (2) Otherwise in compliance with the NIST AI RMF / other recognized framework
Limits for Rebuttable Presumption & Affirmative Defense
• Not mandatory Permissive No timetable
Topics
1. Documentation and requirements for developers
2. Contents of and requirements for notices
3. Content and requirements for risk management policy and program
4. Content and requirements of impact assessments
5. Requirements for rebuttable presumptions
6. Requirements for affirmative defense
AI Inventory
• Understand current and considered uses
Contracts
• Negotiating terms between developers and deployers (e.g., indemnification, data usage/ownership)
Risk Management Program
• Map practices to a framework
• Assessing uses (e.g., impact assessments)
Utah (SB 149)
• Disclosure obligations for use of generative AI
• Effective May 1, 2024
California
• Currently considering ten AI-related bills
• Summary of bills will be published on Byte Back after webinar