This bill, known as the Responsible Artificial Intelligence Security for Employees Act, restricts how employers in Michigan can use automated decision-making tools and electronic monitoring systems for their workforce. It generally prohibits using these technologies for employment decisions like hiring or firing, except when screening large volumes of applications based on specific job criteria. Employers are only allowed to monitor employees for specific purposes such as ensuring job function completion, monitoring production quality, or protecting health and safety, and they must obtain written consent from every employee before using such tools. The legislation also mandates that employers provide clear notice of monitoring, maintain accurate data, allow employees to correct their records, and conduct impact assessments to ensure these tools do not negatively affect individuals based on protected characteristics.
This bill establishes an artificial intelligence governing board and creates a pilot program for state departments and agencies to test generative AI tools. The governor-appointed board will include experts in technology, ethics, and the private sector to develop ethical guidelines and advise on the program's implementation. Starting in January 2027, the pilot program allows approved state employees to use AI tools while prohibiting activities that violate privacy laws, discriminate, or compromise data security. The department of technology, management, and budget will manage the program, including auditing AI usage and providing training to ensure compliance with state and federal standards.
SB 760, the "Leading Ethical AI Development for Kids Act," prohibits operators of companion chatbots from making these AI systems available to minors (under 18) if they could encourage self-harm, unsafe behavior, or harmful interactions like sexual content. The bill specifically bans chatbots that simulate emotional relationships by retaining personal data, asking unsolicited emotional questions, or prioritizing user validation over safety. Operators face $25,000 fines per violation and can be sued by harmed minors or their guardians for damages. This applies to business-owned chatbots designed for ongoing emotional engagement - not customer service or internal tools - and takes effect in 2027.
HB 4048 adds a new sentencing guideline for distributing intimate deepfake sexual images with aggravating circumstances, classifying it as a Class F felony punishable by up to 3 years in prison. This directly affects individuals convicted of disseminating such deepfakes, particularly those with prior offenses under Michigan law. The bill amends sentencing guidelines in Chapter XVII of the Michigan Penal Code to specifically address this crime, establishing a clear penalty framework for courts to apply. It became effective immediately upon the Governor's approval on August 26, 2025.
HB 4668, the "Artificial Intelligence Safety and Security Transparency Act," requires large AI developers (defined as companies spending at least $5 million on a single model or $100 million total in the past year) to implement detailed safety protocols for foundation models. These protocols must address critical risks like mass casualties, $1 billion+ property damage, or AI-enabled weapons, including testing procedures, incident response plans, and physical/digital security measures. Developers must publicly publish safety protocols by 2026 and release quarterly transparency reports detailing risk assessments, critical risk thresholds, and any incidents. The law directly affects major AI companies building advanced systems, mandating concrete safety steps rather than vague guidelines.
