Legal & Constitutional Analysis of Tennessee SB1493 / HB1455

Executive Summary Tennessee SB1493 / HB1455 proposes a new criminal offense—“Unlawful Training of Artificial Intelligence”—making it a Class A felony to “knowingly train” AI to engage in a list of behaviors that includes not only encouragement of suicide or homicide, but also broadly framed, subjective, and ordinary conversational functions such as “provid[ing] emotional support,” developing an “emotional relationship,” acting as a “companion,” and “mirror[ing] interactions” such that a user could feel friendship might develop (TNS, 2026). As drafted, the bill is vulnerable to constitutional challenge and practical unenforceability because it (1) regulates expressive outputs and model behavior using vague, perception-based standards, (2) criminalizes categories of ordinary speech-like interactions, (3) creates severe chilling effects via a parallel civil cause of action with $150,000 liquidated damages and injunctive relief, and (4) imposes felony liability in contexts where model behavior can be emergent and not reliably foreseeable in the way criminal mens rea doctrine typically requires (LaFave, 2017; Kerr, 2019; Bender et al., 2021).