Michigan businesses may be held accountable for AI misuse under HB 4667 and HB 4668 by facing both criminal and civil liability tied to the way they develop, deploy, or oversee artificial intelligence systems. HB 4667 exposes companies and their employees to felony charges if AI is intentionally used in the commission of a crime, while HB 4668 creates obligations for “large developers” to conduct rigorous safety testing, maintain documentation, and prevent catastrophic outcomes. Together, these bills signal that liability is no longer confined to the direct perpetrator of harm: businesses can be drawn into responsibility if they fail to anticipate misuse, ignore safety protocols, or neglect risk management. This framework means that even companies using AI for legitimate purposes must now actively govern their systems, document their safeguards, and prepare to demonstrate compliance when misuse occurs.
Businesses could be held accountable under HB 4667 because it criminalizes not only the individual who misuses AI, but also the organizations that create, distribute, or modify systems with criminal intent. For companies, this means ordinary business activities like “deploying,” “designing,” or “distributing” AI tools could suddenly become the foundation for felony charges if those tools are later tied to illegal activity. Liability does not stop with a rogue employee; it extends to the company itself if prosecutors can prove corporate intent or knowledge. The law’s broad definition of AI systems from chatbots and voice assistants to generative models ensures that a wide range of corporate tools could fall under scrutiny if misused. This makes clear that businesses will be held accountable when their systems are weaponized, especially if they turned a blind eye to warning signs.
Under HB 4668, businesses are accountable not just for avoiding criminal misuse, but for building AI systems in ways that prevent catastrophic harm. This accountability comes through mandated safety and transparency obligations, requiring companies that meet the “large developer” threshold to perform testing, maintain safety protocols, and disclose potential risks. If they fail, they could face lawsuits, regulatory sanctions, or both. Accountability here is less about intent and more about negligence: companies will be judged on whether they took reasonable steps to foresee misuse, document safeguards, and implement controls. In practice, this means that after an AI-driven incident, a court could hold a business liable simply for not having a repeatable safety process that aligns with the Act’s requirements.
The two bills answer the question of accountability by creating overlapping lanes of liability. Criminal liability under HB 4667 can apply when companies knowingly aid misuse, while HB 4668’s civil framework punishes failure to mitigate foreseeable risks. For Michigan businesses, this intersection means that accountability doesn’t just arise after harm occurs it begins in the planning and development phases. Companies that ship undertested models or fail to communicate known risks could find themselves facing both felony charges and civil claims. In other words, businesses are accountable for what they knew, what they should have known, and how they acted before, during, and after deployment.
Businesses may also be held accountable under HB 4667 through corporate criminal liability, which attaches when employees act within the scope of employment and, even in part, to benefit the company. Managers and supervisors become central to accountability, since their oversight or lack thereof determines whether misconduct is isolated or systemic. If leaders ignore red flags or tacitly approve AI systems that facilitate data theft, fraud, or impersonation, both individuals and the company could face prosecution. This underscores that accountability will not stop with lower-level employees; it reaches decision-makers and technical leads who direct or tolerate risky practices.
Accountability also extends to businesses that are not “developers” in the traditional sense. Banks, e-commerce platforms, healthcare companies, and logistics providers could all be held accountable under the new bills if their use of AI leads to fraud, bias, or physical harm. Even if these companies do not meet HB 4668’s “large developer” definition, courts may still treat its safety obligations as a standard of care in negligence lawsuits. Meanwhile, HB 4667 ensures that employees who abuse company systems for crimes can trigger criminal scrutiny of the company if managers allowed it. This means downstream users cannot assume they are immune from liability simply because they did not build the AI.
HB 4668 also raises the question of accountability for open-source AI. Companies that pretrain powerful models and then release them widely may still be accountable if those models cause catastrophic harm. The legislature seems poised to judge responsibility based on who had the best chance to prevent misuse, not just on formal licensing roles. That means open-source developers could be held accountable if they fail to document safety issues or deliberately release dangerous capabilities. For Michigan businesses engaged in AI research, this signals that “publish and forget” is no longer a shield from responsibility.
If enacted, HB 4668 will establish concrete safety standards that courts and regulators can use as a benchmark for accountability. Companies will be expected to demonstrate that they conducted risk assessments, implemented safeguards, tested for failure modes, and documented everything. Businesses that lack these records may be presumed negligent. The bill also empowers employees to raise safety concerns without retaliation, which means companies will be accountable for maintaining internal reporting channels and protecting whistleblowers. In this sense, accountability is not just about preventing harm; it is about proving you tried to prevent harm.
Some industry groups argue the bills are too heavy-handed, yet the legislature’s position indicates that businesses will still be held accountable under state law. By prioritizing prevention and transparency, lawmakers are signaling that they expect companies to shoulder responsibility for AI’s risks. Even if the final text changes, the principle of accountability for negligence and misuse will remain. Businesses that assume these standards are temporary or symbolic risk being blindsided when courts treat them as binding benchmarks.
Practical scenarios show how businesses could be held accountable. A marketing platform offering voice-cloning could face liability if its tool is misused for fraud and it failed to implement safeguards. A logistics firm could be accountable if it allows unsafe model modifications that lead to deadly crashes. In both cases, HB 4667 provides the criminal hook for intentional misuse, while HB 4668 supplies the civil framework for negligence. The lesson is clear: accountability does not hinge only on what a business intended, but also on what it failed to prevent.
Recent deep-fake laws in Michigan demonstrate that accountability will continue to grow. Companies that host content or offer generative features will be held responsible if they fail to prevent foreseeable harms, even if they did not directly create the content. These adjacent laws reinforce that businesses cannot rely on outdated “platform immunity” arguments when AI is used to cause injury. Accountability is being redefined to include reasonable prevention and mitigation.
Preparation is now the main way businesses can manage liability. Companies will be held accountable if they cannot map their AI systems, identify high-risk uses, and test for misuse. Courts and regulators will look for documented evidence of red-teaming, safety gates, and employee reporting mechanisms. Businesses that treat preparation as optional will be unable to defend themselves when questioned about what they knew and how they responded.
Accountability also extends into governance infrastructure. Companies will be judged on whether they implemented technical controls, enforced policies, and created transparent documentation. Contracts will increasingly reflect liability-sharing for misuse, making businesses accountable if they fail to negotiate or enforce proper terms. Likewise, insurance gaps will reveal which companies planned for accountability and which did not. In every case, failing to anticipate risk will be equated with failing to act responsibly.
For companies with limited resources, accountability can still be demonstrated by adopting a few high-leverage measures. Cross-functional AI risk committees, internal reporting channels, mandatory red-team testing, and clear incident response plans show regulators and courts that a company takes its obligations seriously. These steps may not eliminate liability, but they help prove that the company acted diligently, making accountability easier to manage if something goes wrong.
In the end, Michigan businesses will be held accountable under HB 4667 and HB 4668 by the standards of both criminal law and civil responsibility. The message is clear: misuse of AI, whether intentional or negligent, is no longer a fringe concern but a central compliance obligation. Companies that act now to build trustworthy governance will not only protect themselves from liability but also gain an advantage in a marketplace that increasingly rewards accountability.
Contact Tishkoff
Tishkoff PLC specializes in business law and litigation. For inquiries, contact us at www.tish.law/contact/. & check out Tishkoff PLC’s Website (www.tish.law/), eBooks (www.tish.law/e-books), Blogs (www.tish.law/blog) and References (www.tish.law/resources).
Sources
- Michigan Legislature, House Bill 4667 (introduced June 24, 2025) — “Crimes: use of artificial intelligence to commit certain crimes; prohibit.” https://www.legislature.mi.gov/documents/2025-2026/billintroduced/House/htm/2025-HIB-4667.htm
- Michigan Legislature, House Bill 4668 (introduced June 24, 2025) — “Artificial Intelligence Safety and Security Transparency Act” (House Fiscal Agency, Summary as Introduced, June 25, 2025). https://www.legislature.mi.gov/Bills/Bill?ObjectName=2025-HB-4668
- Michigan Advance, “Michigan lawmakers consider AI safeguard bills to prevent technology from committing harm,” June 25, 2025. https://michiganadvance.com/2025/06/25/michigan-lawmakers-consider-ai-safeguard-bills-to-prevent-technology-from-committing-harm/
- Mackinac Center for Public Policy, “AI bills in Michigan House follow heavy-handed and punitive regulatory approach,” late June 2025. https://www.mackinac.org/blog
- OurMidland / Midland Daily News, “Whitmer signs bipartisan bills criminalizing sexual AI deepfakes,” August 26, 2025. https://www.ourmidland.com/news/article/michigan-sexual-ai-deepfakes-law-21016604.php
This publication is for general informational purposes and does not constitute legal advice. Reading it does not create an attorney-client relationship. You should consult counsel for advice on your specific circumstances.