Minnesota’s chatbot liability bill is a constitutionally suspect speech regulation that broadly restricts entire categories of lawful AI-generated content without clear limiting principles, making it vulnerable to First Amendment challenge. The bill’s vague standards, strict liability provisions and expansive definition of “chatbot” impose unworkable technical mandates that will force companies to over-censor or withdraw beneficial tools.
NetChoice Testimony in Opposition to Minnesota SF 4997 and Proposed Amendments
April 24, 2026
Minnesota Legislature
Minnesota Senate Judiciary and Public Safety Committee
Dear Chair Latz, Vice-Chair Verbeten, Ranking Member Limmer and Members of the Committee:
NetChoice respectfully requests your opposition to SF 4997, which imposes sweeping liability and operational restrictions on artificial intelligence chatbot technologies, including so-called “companion chatbots.”
We share the legislature’s goal of protecting users—especially minors—from harm. But this proposal, while well-intentioned, adopts an approach that is constitutionally suspect, operationally unworkable and likely to produce consequences that undermine both innovation and user safety. Rather than addressing specific harms, the bill creates a broad liability regime that will chill lawful speech, expose companies to expansive litigation risk and incentivize the withdrawal of beneficial AI tools from Minnesota consumers.
The Bill Imposes Unconstitutional Restrictions on Protected Speech
At its core, this legislation prohibits chatbots from generating certain categories of “substantive” information or advice—specifically, content that would require a professional license if delivered by a human. This is a direct regulation of speech. Courts have consistently held that the First Amendment protects not only human speakers, but also the dissemination of information through technological intermediaries.
Chatbots today are used for a wide range of lawful and socially valuable expression: educational support, general health information, legal background explanations and everyday problem-solving. The bill’s prohibition sweeps far beyond harmful conduct and instead restricts entire categories of speech based on content. Critically, it does so without clear limiting principles—what constitutes “substantive advice” is undefined and inherently subjective.
Moreover, the bill explicitly rejects disclaimers as a mitigating measure. This eliminates a widely accepted and less restrictive alternative, further weakening its constitutional footing. Courts evaluating speech restrictions routinely require governments to adopt narrower approaches when available. This bill does the opposite.
The Liability Framework is Expansive and Unworkable
The legislation creates a private right of action with damages, attorney’s fees and enhanced penalties for willful violations. At the same time, it establishes vague and technically complex obligations that no developer can reliably operationalize.
For example, chatbot operators must ensure their systems do not provide outputs that could be construed as regulated professional advice. But modern AI systems generate probabilistic, context-dependent responses. Drawing a bright line between permissible information and prohibited “advice” is not a solvable engineering problem. Faced with this uncertainty and asymmetric litigation risk, companies will be forced to over-censor outputs or withdraw services entirely.
The bill also imposes strict liability in certain contexts, particularly involving minors and self-harm. Strict liability regimes are typically reserved for inherently dangerous activities with clear causal chains. Here, however, liability attaches to complex, multi-factor human behavior, often mediated through conversational systems that do not control user actions. This creates an untenable legal environment in which companies are held responsible for outcomes they cannot reasonably predict or prevent.
Companion Chatbot Provisions Mandate Intrusive Monitoring and Restrict Access to Helpful Tools
The bill’s treatment of “companion chatbots” is especially problematic. It requires companies to detect user mental states, including expressions of self-harm, and to intervene by suspending access and directing users to crisis resources.
While the goal is understandable, the mechanism raises serious concerns. To comply, companies would need to continuously monitor and analyze the content of user conversations at a granular level. This level of surveillance implicates significant privacy concerns and may deter users—particularly vulnerable individuals—from seeking support or information in the first place.
Additionally, the mandatory 72-hour lockout requirement is rigid and disconnected from clinical realities. It removes user autonomy and may cut off access to tools that, in many cases, provide comfort, structure or de-escalation. A one-size-fits-all intervention model is ill-suited to nuanced mental health contexts.
Age Determination Requirements Create New Privacy and Security Risks
The bill requires chatbot operators to make “good faith” efforts to determine whether users are minors, with strict liability consequences for failure. In practice, this will push companies toward implementing age verification or estimation systems.
These systems often rely on sensitive personal data, including identification documents or biometric inference. Creating and maintaining such systems introduces significant cybersecurity risks. A database used to verify age—even temporarily—becomes a high-value target for malicious actors.
The result is a paradox: in attempting to protect minors, the bill encourages the collection of precisely the kinds of sensitive data that increase their exposure to harm.
The Bill’s Definitions are Overbroad and Capture Mainstream Technologies
The definition of “chatbot” is expansive, encompassing any system that simulates conversation. This includes not only companion-style applications but also customer service tools, educational platforms, productivity assistants and general-purpose AI systems.
By failing to narrowly tailor its scope, the bill subjects a wide array of benign and beneficial technologies to the same liability framework designed for a narrow set of concerns. This overbreadth will stifle innovation and create uncertainty across multiple sectors that rely on conversational interfaces.
The combination of vague standards, strict liability in key areas and fee-shifting provisions creates strong incentives for litigation. This will not be limited to bad actors. Compliant companies—particularly smaller developers—will face significant pressure to settle even meritless claims due to the cost and unpredictability of litigation. The likely outcome is a “litigation tax” on innovation, rather than meaningful improvements in user safety.
This bill reflects legitimate concerns about emerging technologies, but it adopts a blunt and legally vulnerable approach. By regulating speech, imposing unworkable technical mandates and creating expansive liability exposure, it risks reducing access to beneficial tools while failing to effectively target harmful conduct. The proposed amendments from S.F. 1858 and S.F. 4696 would compound these defects substantially.
We respectfully urge the legislature to oppose this bill and any amendments to incorporate S.F. 1858 or S.F. 4696. We instead urge the pursuit of approaches that are narrowly tailored, constitutionally sound and grounded in technical feasibility. NetChoice stands ready to work with policymakers to develop solutions that protect users without undermining innovation or fundamental rights.
Amy Bos
Vice President of Government Affairs, NetChoice (The views of NetChoice expressed here do not necessarily represent the views of all NetChoice members.)
NetChoice is a trade association that works to protect free expression and promote free enterprise online.