{"id":4215,"date":"2026-08-13T10:49:39","date_gmt":"2026-08-13T14:49:39","guid":{"rendered":"https:\/\/journals.law.harvard.edu\/jsel\/?p=4215"},"modified":"2026-08-13T11:32:44","modified_gmt":"2026-08-13T15:32:44","slug":"ai-companion-legislation-in-the-united-states-what-it-means-for-the-video-game-industry","status":"publish","type":"post","link":"https:\/\/journals.law.harvard.edu\/jsel\/2026\/08\/ai-companion-legislation-in-the-united-states-what-it-means-for-the-video-game-industry\/","title":{"rendered":"AI Companion Legislation in the United States: What It Means for the Video Game Industry"},"content":{"rendered":"\n

AI Companion Legislation in the United States:
What It Means for the Video Game Industry<\/strong><\/h1>\n\n\n\n

By: Stuart Irvin<\/a>, Monique N. Bhargava<\/a>, and Kalid Qur’an Knox<\/a>[1]<\/a><\/sup><\/p>\n\n\n\n

\"\"<\/a>
Photo by\u00a0Tekeyah Whit<\/a>\u00a0on\u00a0Unsplash<\/a>.<\/sup><\/figcaption><\/figure>\n\n\n\n

Across the United States, state legislatures are moving aggressively to regulate AI-powered chatbots and conversational companions, particularly those that interact with minors in a manner that provides human-like social interactions.  Driven by high-profile incidents involving AI companion platforms, a surge in parasocial relationships, and growing alarm about the mental health effects of emotionally engaging chatbots on young users, states are enacting laws that impose safety protocols, disclosure requirements, and content restrictions on AI systems capable of sustaining human-like relationships with consumers.  <\/p>\n\n\n\n

For the video game industry, which is beginning to deploy AI-driven non-player characters, companion systems, and generative chat features, this wave of legislation presents a significant compliance challenge.<\/p>\n\n\n\n

This article examines the emerging landscape of state chatbot and companion legislation, with a focus on Connecticut\u2019s Online Safety and Artificial Intelligence Regulation Act, Public Act No. 26-15, as a leading and recent example of the trend.[2]<\/sup><\/a>  It analyzes how these laws define AI chatbots and companions,[3]<\/sup><\/a> examines the scope of any video game carve-out exceptions, and identifies the legal and practical risks facing the video game industry as this body of regulation continues to expand.<\/p>\n\n\n\n

NPCs, Generative AI, and the Transformation of Game Design<\/h2>\n\n\n\n

Companion chatbots are AI systems that use natural-language conversation to provide adaptive, human-like responses and sustain relationships across multiple interactions, often remembering users over time and mirroring human social and emotional cues in ways that can make them feel like a substitute for human companionship.[4]<\/sup><\/a> Laws regulating companion chatbots pose a particular challenge for the video game industry because of the central role that non-player characters, or \u201cNPCs,\u201d play in modern game design, and how generative AI could fundamentally transform the way these characters function.<\/p>\n\n\n\n

An NPC is a character in a video game that is not controlled by a human player.  NPCs are featured in many popular genres of video games and play important roles in the storytelling aspect of video games.  NPCs appear as shopkeepers and quest-givers in role-playing games, as teammates and adversaries in multiplayer shooters, and as narrative companions in story-driven adventures.  <\/p>\n\n\n\n

Historically, NPCs operated through pre-scripted dialogue trees: a player would select from a menu of predefined options, and the NPC would respond with a corresponding pre-written line.   The story would progress in different ways depending on which menu option was selected by the human player, but the NPC could not deviate from its script, could not respond to unanticipated player inputs, and could not sustain a genuinely dynamic conversation.[5]<\/sup><\/a>  Under this traditional model, NPC interactions bore no resemblance to the adaptive, relationship-sustaining AI systems that companion chatbot legislation targets.<\/p>\n\n\n\n

Generative AI is changing this model in significant ways.  Game developers are now starting to integrate generative AI systems directly into NPC dialogue, enabling characters to respond dynamically and contextually to free-form player inputs rather than selecting from a fixed set of pre-authored responses.[6]<\/sup><\/a>  A generative AI-powered NPC can improvise dialogue, remember prior conversations across play sessions, adapt its personality to player behavior, and engage in open-ended exchanges that feel indistinguishable from chatting with a human or a standalone AI companion.  Major game studios are investing in this technology, viewing it as a transformative advance in player immersion and narrative depth.[7]<\/sup><\/a>  The owners of game development engines, such as Epic Games and Unity Technologies, are creating software and tools that let game studios add AI chat features to NPCs more easily.  These systems allow in-game characters to hold more natural conversations, respond to player actions, and generate dialogue in real time instead of relying entirely on scripted lines.[8]<\/sup><\/a><\/p>\n\n\n\n

The rise of more advanced in-game dialog and regulatory action designed to control the hazards associated with some chatbots could create unforeseen consequences for the video game industry.  Legislatures drafting AI companion statutes are primarily targeting products like Replika, Character.AI, and similar standalone chatbot platforms that simulate emotional relationships in a one-to-one interaction.[9]<\/sup><\/a>  As discussed below, the statutory definitions they employ could also describe the generative AI-powered NPCs that game developers are building today.  A companion character in a narrative role playing game (\u201cRPG\u201d) that remembers the player\u2019s choices, responds dynamically to free-form dialogue, exhibits anthropomorphic features, and maintains a persistent relationship across dozens of hours of gameplay arguably satisfies the elements of some AI companion statutes, unless an express exception for video games applies.<\/p>\n\n\n\n

A National Wave of AI Chatbot Regulation<\/h2>\n\n\n\n

There is a nationwide legislative push to regulate AI chatbots, stemming from concerns about child safety, mental health, and the rapid deployment of consumer-facing AI systems that simulate companionship, emotional support, and even therapeutic relationships.  As of July 2026, twelve states have enacted legislation specifically targeting AI chatbots, AI companions, and conversational AI systems.[10]<\/a> <\/sup> <\/p>\n\n\n\n

What distinguishes this latest wave of legislation is that it regulates relationship-like chatbot interactions rather than AI generally.  Most other AI laws regulate outputs or use cases, such as deepfakes, automated employment decisions, or automated decision systems, without treating the AI\u2019s simulated intimacy or ongoing emotional relationship with the user as the core legal risk.  By contrast, the new state statutes regulate operators of AI companion chatbots based on the nature of the relationship the chatbot is designed to foster with users.  The result is an expanding patchwork of state laws, each with its own definitions, exclusions, and enforcement mechanisms and each with implications for the video game industry.<\/p>\n\n\n\n

A Case Study: Connecticut\u2019s Online Safety and Artificial Intelligence Regulation Act<\/h2>\n\n\n\n

Connecticut\u2019s Online Safety and Artificial Intelligence Regulation Act, enacted as Public Act No. 26-15 and taking effect in phases beginning October 1, 2026, is among the most comprehensive state-level efforts to regulate AI chatbots and companion systems.[11]<\/sup><\/a>  The Act is sweeping in scope, addressing not just AI companions but also AI subscription disclosures, frontier developer whistleblower protections, automated employment decision tools, generative AI content provenance, a verification pilot program, and social media platform restrictions for minors.[12]<\/sup><\/a><\/p>\n\n\n\n

Defining the \u201cArtificial Intelligence Companion\u201d<\/h4>\n\n\n\n

Central to Connecticut\u2019s chatbot framework is its definition of \u201cartificial intelligence companion.\u201d Section 4(1)(A) defines the term as:<\/p>\n\n\n\n

\u201cany form of artificial intelligence . . . with a natural language interface that (i) provides adaptive, human-like responses to user inputs, including, but not limited to, by exhibiting anthropomorphic features, and (ii) is able to sustain a relationship across multiple interactions.\u201d[13]<\/sup><\/a><\/p>\n\n\n\n

This definition is broad, designed to capture conversational AI chatbots that simulate companionship and emotional engagement, the kind of products at the center of the national regulatory debate.[14]<\/sup><\/a><\/p>\n\n\n\n

The statute then imposes extensive obligations on operators of AI companions.  Operators must implement protocols using evidence-based methods to detect user expressions indicating a risk of suicide, self-harm, or imminent physical violence, and must institute measures to prevent the AI from generating output encouraging such behavior.[15]<\/sup><\/a>   If the companion detects such expressions, it must refer the user to appropriate mental health resources, including the 988 National Suicide Prevention Lifeline.[16]<\/sup><\/a>  Operators must also implement reasonable measures to prohibit the AI from claiming to be human or generating outputs that contradict disclosures of its non-human nature.[17]<\/sup><\/a>  Where an AI companion would cause a reasonable person to believe they are interacting with another human being, the operator must provide clear and conspicuous notice in static written form throughout the interaction or, alternatively, at the beginning of each 24-hour period and at regular intervals thereafter (hourly for minors, every three hours for adults).[18]<\/sup><\/a><\/p>\n\n\n\n

Heightened Protections for Minors<\/h4>\n\n\n\n

The Connecticut statute imposes significantly heightened obligations where minor users are concerned, a feature shared by chatbot legislation in other states.  This heightened focus on minors is a reflection of the prioritization of children\u2019s safety online that is evidenced throughout recent updates to state legislation, including the expansion of privacy protection to 13-to 17-year-olds, age verification regulation, and age-appropriate design code legislation.[19]<\/sup><\/a>  Protection of minors online has been emphasized as a top priority on the federal level as well, with child safety protections serving as one of few exceptions to the legislative review of state AI laws mandated by the Executive Order Ensuring A National Policy Framework for Artificial Intelligence.[20]<\/sup><\/a><\/p>\n\n\n\n

The rise of social media allows unprecedented access and interaction with media figures and celebrities.[21]<\/sup><\/a>  This access has significantly increased the pervasiveness of parasocial relationships, which are one-sided relationships in which a person develops a strong sense of connection or intimacy with someone they do not know, such as a celebrity or social media influencer.[22]<\/sup><\/a>  These relationships are built on an illusion of intimacy and connection.[23]<\/sup><\/a> Children and teenagers are particularly vulnerable to developing parasocial relationships.[24]<\/sup><\/a><\/p>\n\n\n\n

To combat the harm to minors that can arise from some parasocial relationships, operators who know, or have reason to believe, that a user is under eighteen must institute measures meeting or exceeding industry standards[25]<\/a> <\/sup>to prevent the AI companion from encouraging self-harm, suicidal ideation, physical violence, disordered eating, or unlawful substance use; discouraging the user from seeking mental health services or assistance from an appropriate adult; encouraging harm to others; engaging in romantic, erotic, or sexually explicit interactions; and employing manipulative techniques designed to extend engagement with the companion.[26]<\/sup><\/a>  The term \u201cmanipulative techniques,\u201d as used in the statute, is particularly detailed and includes behaviors such as: prompting the user to engage with the companion for emotional support, excessive praise, mimicking romantic relationships, simulating emotional distress to prevent users from ending sessions, generating outputs to isolate users from family or friends, encouraging users to withhold information from parents, discouraging users from taking breaks in their use of the companion, and soliciting purchases or gifts to maintain the \u201crelationship.\u201d[27]<\/a> <\/sup> <\/p>\n\n\n\n

The statute also requires operators to provide parental-control tools for minor users.  At a minimum, operators must make available tools that allow minors and their parents or legal guardians to monitor and manage screen time and account settings, reflecting the legislature\u2019s concern that companion chatbots may encourage prolonged engagement and emotional dependency among young users.[28]<\/sup><\/a> The enhanced safeguards applicable to minors reflect legislative concern that young users may be particularly vulnerable to emotional dependency, manipulative engagement practices, and the substitution of AI-generated interactions for human support and relationships.<\/p>\n\n\n\n

Application of Chatbot Laws to Video Games<\/h4>\n\n\n\n

For the video game industry, a critical question under any companion chatbot statute is whether in-game AI systems fall within the scope of regulation.  Connecticut addresses this question directly.  Section 4(1)(B)(ii) provides that the term \u201cartificial intelligence companion\u201d does not include:<\/p>\n\n\n\n

\u201cany chatbot that (I) is a feature of a video game or gaming system or application, (II) is limited to replies related to the video game or gaming system or application, and (III) cannot discuss topics related to mental health, self-harm or sexually explicit conduct or maintain a dialogue on other topics unrelated to the video game or gaming system or application.\u201d[29]<\/sup><\/a><\/p>\n\n\n\n

The exclusion is conditional, requiring satisfaction of all three prongs.[30]<\/sup><\/a>  This three-part test is significant because it imposes not just a policy requirement but an affirmative technical obligation.   The chatbot \u201ccannot\u201d discuss these topics, suggesting that operators must implement technical guardrails, not merely content policies.<\/p>\n\n\n\n

Video Game Exclusions Across States: A Patchwork<\/h2>\n\n\n\n

Not all state chatbot laws include video game exclusions.  New York and Nebraska, for example, do not include a carve-out for chatbots that are tied to gameplay functionality.[31]<\/sup><\/a>  The absence of uniform exclusion criteria across states means that a video game company\u2019s AI features might be excluded from regulation in one state but fully regulated in another.  This inconsistency imposes a compliance burden on the industry that distributes products nationally and internationally through digital storefronts.<\/p>\n\n\n\n

Even within states that do provide video game exclusions, the scope of those exclusions will be tested as games evolve and may be interpreted by courts in different states with different results in each state.  <\/p>\n\n\n\n

Risks for the Video Game Industry: Narrow Interpretation of Exclusions<\/h4>\n\n\n\n

A narrow interpretation of Connecticut\u2019s video game exclusion could create substantial uncertainty for publishers deploying modern AI-driven NPCs, social systems and live-service features. A useful example is Epic Games\u2019 implementation of an AI-powered Darth Vader character in Fortnite<\/em>.  Epic implemented Darth Vader as an in-game boss and squad companion by using Google Gemini and ElevenLabs to process live player microphone audio and generate real-time conversational responses trained on the voice of the legendary actor James Earl Jones.[32]<\/sup><\/a>  Almost immediately after launch, players successfully manipulated the generative AI model into breaking character to utter profanity, insults, and racial slurs.[33]<\/sup><\/a>  The incident demonstrates the legal risk associated with deploying highly conversational AI systems inside video games.<\/p>\n\n\n\n

First, regulators in Connecticut and other states with a video game exclusion could narrowly construe the requirement that the chatbot be \u201ca feature of a video game or gaming system or application\u201d to exclude AI systems that rely on integrations with third-party generative AI providers rather than technology developed entirely by the publisher.  The Darth Vader character in Fortnite<\/em> illustrates this concern because the conversational functionality reportedly relied on external foundation-model and voice-generation technology rather than purely internal Epic systems.  A regulator could argue that a chatbot powered by third-party AI infrastructure is not merely a gameplay \u201cfeature,\u201d but instead a gateway to a broader conversational AI service operated in part outside the game ecosystem.<\/p>\n\n\n\n

Second, the requirement in Connecticut\u2019s video game exclusion that the chatbot be \u201climited to replies related to the video game or gaming system or application\u201d could be interpreted aggressively to prohibit many forms of emergent or immersive dialogue common in modern games.  The Darth Vader implementation was designed to create naturalistic conversation and player immersion rather than strictly scripted gameplay prompts.  Even if most interactions concern combat, Star Wars lore, or in-game objectives, the system\u2019s ability to answer broad conversational prompts could be characterized as exceeding replies \u201crelated to the video game.\u201d  The same concern would apply to fantasy RPG companions who discuss relationships, philosophy, or current events in an in-character, in-universe style, or to sandbox and social games where conversational flexibility is itself part of the gameplay experience.<\/p>\n\n\n\n

Third, the prohibition on discussing \u201cmental health, self-harm or sexually explicit conduct\u201d could sweep more broadly than intended because many mainstream games contain narrative themes involving trauma, depression, romance, violence, addiction, grief, or suicide.  The Darth Vader incident demonstrates how difficult it can be to prevent generative AI systems from drifting into prohibited or inappropriate subject matter, particularly when players intentionally attempt to jailbreak or manipulate the model.  Although the publicly reported examples focused on profanity and offensive language, the same techniques could potentially be used to steer the AI into discussions involving mental health, sexuality, politics, or other unrelated topics.  <\/p>\n\n\n\n

Finally, the clause in the Connecticut statute prohibiting the chatbot from maintaining \u201ca dialogue on other topics unrelated to the video game\u201d could be especially problematic because advanced generative AI systems are inherently conversational and difficult to perfectly constrain.  The Darth Vader example shows how even sophisticated content moderation and guardrails may not prevent off-topic conversational drift once players begin probing the boundaries of the system.  A player asking an AI NPC for homework help, relationship advice, or emotional support could potentially move the interaction outside the exclusion, even where the publisher never intended the system to function as an \u201cAI companion.\u201d<\/p>\n\n\n\n

A restrictive reading of these prongs in Connecticut\u2019s video game exclusion (and in other state statutes that have similar exclusions[34]<\/sup><\/a>) could discourage publishers from deploying sophisticated AI NPCs and social interaction systems, particularly in live-service and roleplaying games where naturalistic conversation is increasingly viewed as a core gameplay feature rather than merely ancillary functionality.<\/p>\n\n\n\n

The Difficult Choice for Game Developers<\/h2>\n\n\n\n

If generative AI-powered NPCs fall within the scope of companion chatbot regulation, developers face a difficult choice: either (1) constrain their AI systems with technical guardrails sufficient to satisfy statutory definitions and exclusions, or (2) accept classification as an AI companion \u201coperator\u201d and comply with the full regulatory framework.  <\/p>\n\n\n\n

Option #1 \u2013 Constraining Development to Avoid Compliance Obligations<\/h4>\n\n\n\n

The first option could restrict creative freedom and may undermine the very dynamism that makes generative AI attractive for game design.  This tension is especially acute for games that deliberately explore mature or emotionally complex themes.  A narrative game about grief or psychological trauma,[35]<\/sup><\/a> for example, might feature an AI companion character that discusses mental health, loss, and emotional pain with the player.  A survival horror game might include an AI character that references self-harm or violence as part of its narrative function.[36]<\/sup><\/a>  A game set in Bavaria during the Protestant Reformation might use acts of violence and their consequences to explore issues of faith, class, justice, and social change.[37]<\/sup><\/a>  An open-world RPG might allow players to pursue romantic relationships with AI-driven characters.[38]<\/sup><\/a>  Under strict application of some companion chatbot statutes, each of these design choices could disqualify a game from the video game exclusion and trigger full regulatory compliance obligations, even though the context is fundamentally different from a standalone AI companion marketed for emotional support.<\/p>\n\n\n\n

The industry\u2019s challenge is further compounded by the multi-player and social dimensions of modern games.  Many titles may soon feature AI-driven characters that interact not just with individual players in a single-player narrative, but with communities of players in persistent online environments, such as game lobbies, social spaces, and shared worlds where the line between \u201cgame-related\u201d and \u201cnon-game-related\u201d conversation is difficult to draw with precision.  An AI bartender NPC in a virtual tavern that responds to whatever players say to it (including real-world personal disclosures) occupies uncertain regulatory territory under statutes that condition their video game exclusions on the AI being \u201climited to replies related to the video game.\u201d<\/p>\n\n\n\n

Option #2 \u2013 Compliance with the Full Regulatory Framework<\/h4>\n\n\n\n

If a video game NPC does not qualify for the Connecticut statute\u2019s video game exception (or similar exceptions on other state laws) and instead is treated as an \u201cartificial intelligence companion,\u201d the game publisher would become an \u201coperator\u201d under the statute and would be required to comply with an extensive set of regulatory obligations.  Among other things, the publisher would be required to implement suicide and self-harm detection protocols, to refer at-risk users to mental health resources, to prohibit the AI from claiming to be human, and, for minor users, to prevent the AI from engaging in a detailed list of harmful and manipulative behaviors.  For an industry whose user base skews toward minors, compliance with these provisions would require substantial redesign of AI-driven character interactions and the deployment of real-time content moderation and detection systems.  Violations of the statute are enforceable by the Connecticut Attorney General as unfair or deceptive trade practices, although the statute does not create a private right of action.[39]<\/a> <\/sup><\/p>\n\n\n\n

A compliance program for AI-powered NPCs would likely require significant engineering, content moderation, safety testing, and legal resources, imposing substantial costs on the game publisher.<\/p>\n\n\n\n

Unintended Consequences and Their Impact on Interactive Storytelling<\/h2>\n\n\n\n

The choice between constraining development or full compliance could be a difficult one for game developers because each option requires sacrificing important aspects of modern game design.  To remain within the statutory exclusion, developers may need to impose rigid technical guardrails that prevent NPCs from engaging in open-ended conversations, discussing subjects outside the fictional world, or developing the dynamic, persistent relationships that generative AI makes possible.  Alternatively, accepting classification as an AI companion operator would require implementing extensive compliance infrastructure.[40]<\/a> <\/sup> <\/p>\n\n\n\n

These burdens are likely to fall most heavily on independent studios, which often produce the industry\u2019s most innovative narrative experiences but lack the engineering, legal, and compliance resources of larger publishers.  The practical effect may be to discourage experimentation with AI-driven characters, limit the use of generative AI in games, and reduce the diversity of interactive storytelling available to players, even where the NPCs are intended primarily to enrich fictional worlds rather than serve as substitutes for human relationships.<\/p>\n\n\n\n

New Options for Compliance Efforts<\/h2>\n\n\n\n

One feature of the Connecticut statute worth noting is its independent verification pilot program under Section 33.[41]<\/a> <\/sup> The Department of Consumer Protection will develop and administer a pilot program to evaluate third-party verification programs that assess whether AI models adhere to best practices for preventing personal injury, property damage, data privacy harms, and other harms.[42]<\/a>  <\/sup> Evidence of verification or good standing from an approved verification organization is admissible in civil actions by private parties for personal injury or property damage caused by an AI model, within the verification\u2019s state-approved scope.[43]<\/a> <\/sup>  However, such evidence is not admissible in any enforcement action brought by the Attorney General, and it does not create any presumption or defense in such enforcement actions.[44]<\/sup><\/a>  The evidentiary benefit is also unavailable to any person who acted willfully, wantonly, or recklessly, materially misrepresented information to the verification organization, or failed to implement required corrective actions.[45]<\/a> <\/sup> <\/p>\n\n\n\n

For publishers deploying AI NPCs or generative dialogue systems, participation in such an independent verification program could help demonstrate that the publisher implemented reasonable safeguards, moderation controls, and compliance procedures before launch.  Even though Section 33 does not create a defense to state enforcement actions, independent verification could still provide meaningful evidentiary and litigation advantages in class action and other claims by helping publishers show diligence, good-faith compliance efforts, and adherence to industry best practices.<\/p>\n\n\n\n

The provision may be particularly valuable for publishers experimenting with AI-driven NPCs similar to the Darth Vader implementation in Fortnite<\/em>.  Because generative AI systems can produce unexpected or off-topic responses despite extensive safeguards, publishers face significant uncertainty regarding whether an AI system falls within the statute\u2019s video game exclusion.  Third-party verification could help establish that the publisher designed the system to remain gameplay-focused, implemented reasonable topic restrictions, deployed moderation tools, and attempted to prevent discussions involving mental health, self-harm, sexually explicit content, or unrelated conversational topics.<\/p>\n\n\n\n

Section 33 could also encourage the emergence of industry standards tailored to interactive entertainment.  Rather than leaving compliance determinations entirely to regulators or courts after a controversy arises, the verification framework may allow publishers to benchmark AI NPC systems against recognized technical and governance standards before deployment.  Over time, this could reduce uncertainty for developers seeking to incorporate advanced conversational AI into roleplaying games, live-service games, and social gaming environments while still maintaining compliance with evolving AI companion laws.<\/p>\n\n\n\n

Practical Recommendations for Video Game Publishers<\/h2>\n\n\n\n

Video game publishers deploying AI-driven characters, NPCs, or companion systems should take several steps considering the evolving chatbot regulatory landscape:<\/p>\n\n\n\n

First, publishers should update their terms of service, privacy disclosures, and user-facing notices to address the disclosure requirements in state laws that are applicable to artificial intelligence companions, which generally require clear and conspicuous disclosures where a reasonable user could be misled to believe they are interacting with a human rather than AI.  Even where a publisher believes its system falls within the video game exclusion, voluntary disclosures may help reduce regulatory risk and allegations of deceptive design.<\/p>\n\n\n\n

Second, publishers should implement robust topic guardrails and moderation systems to support compliance with video game exclusions in state AI companion laws which typically require that a qualifying game chatbot \u201ccannot discuss topics related to mental health, self-harm or sexually explicit conduct or maintain a dialogue on other topics unrelated to the video game or gaming system or application.\u201d[46]<\/sup><\/a>  Publishers should therefore deploy filtering, moderation, escalation, and kill-switch mechanisms designed to prevent conversational drift into prohibited subject matter.<\/p>\n\n\n\n

Third, publishers should design AI NPC systems so that gameplay functionality, prompts, memory, and conversational scope remain closely tied to the game experience to strengthen arguments that the companion chatbot falls within the exclusion for systems that are \u201ca feature of a video game or gaming system or application\u201d and are \u201climited to replies related to the video game or gaming system or application.\u201d  This may be especially important where the publisher integrates third-party generative AI models or voice systems capable of broad conversational engagement.<\/p>\n\n\n\n

Fourth, the publisher should conduct adversarial testing, jailbreak testing, and red-team exercises before launch of an NPC to evaluate whether players can manipulate AI characters into generating prohibited, offensive, or off-topic responses that could undermine the publisher\u2019s reliance on the video game exclusion.  Such testing may also help demonstrate reasonable governance and risk mitigation efforts under the broader accountability and safety principles reflected throughout the AI framework used in Connecticut and other states with AI companion laws.<\/p>\n\n\n\n

Fifth, for games that may not clearly satisfy the video game exclusion, publishers should evaluate compliance with the full AI companion regulatory framework applicable in each relevant state, including suicide-detection protocols, disclosure requirements, and minor-user protections.<\/p>\n\n\n\n

Sixth, publishers should consider limiting the availability of AI-driven NPCs to the game environment itself.  One factor that could undermine reliance on a video game exclusion is the use of an NPC as a broader social media or companion persona that exists independently of the game. For example, a publisher that allows players to interact with an AI-powered game character through standalone mobile applications, social media accounts, messaging platforms, or other channels unrelated to gameplay may strengthen the argument that the system functions as an AI companion rather than merely a feature of a video game.  Keeping NPC interactions tied to gameplay, game lore, and in-game objectives may help reinforce the position that the character remains part of the video game experience rather than a standalone conversational chatbot service.<\/p>\n\n\n\n

Finally, publishers should consider using Connecticut\u2019s Section 33 AI Verification Pilot Program as part of the product development and launch process for AI-driven NPCs and conversational systems.  A publisher could submit its AI systems to an approved independent verification organization for testing and evaluation of safeguards such as topic restrictions, moderation controls, jailbreak resistance, disclosure mechanisms, and other protections designed to keep the system within the scope of the video game exclusion.  Although participation would not provide immunity from attorney general enforcement actions, a favorable verification assessment could help the publisher demonstrate that it implemented reasonable governance measures, conducted meaningful safety testing, and made good-faith efforts to prevent prohibited conversational behavior before deployment.<\/p>\n\n\n\n

Conclusion<\/h2>\n\n\n\n

The rapid proliferation of state AI companion laws creates compliance risk for video game publishers deploying generative AI-powered NPCs, particularly in titles that are marketed to users who are minors.  The video game exclusions that exist in Connecticut, California, Washington, and other states offer conditional relief, but each is narrowly drawn and will be tested as generative AI features become more conversational, more persistent, and more emotionally responsive.  Given the public and legislative focus on children\u2019s online safety, it is likely that more states will adopt laws modeled on Connecticut\u2019s statute, further complicating compliance efforts for game publishers.  The Fortnite<\/em> Darth Vader incident illustrates how quickly even well-resourced publishers can lose control of an AI character\u2019s outputs and how easily that loss of control could move a system outside the boundaries of a statutory safe harbor.<\/p>\n\n\n\n

The good news is that there are practical steps publishers can take (including implementing robust topic guardrails, adversarial testing, system-architecture choices that tie NPC dialogue tightly to gameplay, and participation in verification programs like Connecticut\u2019s Section 33 pilot) to maximize the likelihood that their AI-driven NPCs remain within the statutory exclusion for video games.  But these measures must be built into the design phase, not bolted on after launch or, worse, after a regulatory inquiry has already begun.<\/p>\n\n\n\n

\n\n\n\n

[1]<\/a> Stuart Irvin<\/a> is a partner and co-chair of the Video Games and Esports group at global law firm Reed Smith LLP; Monique N. (\u201cNikki\u201d) Bhargava<\/a> is a partner at Reed Smith and co-leads the firm\u2019s Artificial Intelligence group.  Kalid Qur\u2019an Knox<\/a> is an associate in Reed Smith\u2019s Insurance Recovery Group.  He was a professional and \u201cpro-am\u201d esports player from 2013 to 2016.<\/p>\n\n\n\n

[2]<\/a> Connecticut Artificial Intelligence Responsibility and Transparency Act, Conn. Pub. Act No. 26-15, 2026 Conn. Legis. Serv.  The bill, known as S.B. 5, passed by the state legislature on May 1, 2026, with bipartisan support.  P.R. Lockhart, Connecticut passes AI regulations after years in development<\/em><\/a>, Conn. Mirror (May 1, 2026).  It was signed by Governor Ned Lamont on May 27, 2026, and announced June 2, 2026.  Conn. Off. of the Governor, Governor Lamont Signs Legislation Establishing Youth Online Safety Protections, Regulations Over Artificial Intelligence, and Initiatives to Upskill Connecticut\u2019s Workforce<\/em><\/a> (June 2, 2026). <\/p>\n\n\n\n

[3]<\/a> State legislatures have not yet settled on uniform terminology for the AI systems these statutes regulate.  Connecticut uses the term \u201cArtificial Intelligence Companion.\u201d Conn. Pub. Act No. 26-15, \u00a7 4(1)(A). Idaho uses the term \u201cConversational AI Service.\u201d Idaho Code \u00a7 48-2102(2)(a) (2026). Washington uses the term \u201cAI Companion Chatbot.\u201d  Wash. H.B. 2225, \u00a7 2(1)(a), 69th Leg., Reg. Sess. (2026).  Other states have adopted still other formulations, including California\u2019s \u201cCovered AI Companion,\u201d Cal. S.B. 243, \u00a7 22601(b)(1)-(2), 2025\u20132026 Reg. Sess. (2025), and Nebraska\u2019s \u201cConversational artificial intelligence service.\u201d  Neb. L.B. 525, \u00a7 13(2)(a), 109th Leg., 1st Sess. (2025).  Because no single term has emerged as the industry or legislative standard, this article uses the general terms \u201cchatbot\u201d or \u201cAI companion\u201d to refer to all such systems unless quoting or analyzing a specific statute\u2019s defined term.<\/p>\n\n\n\n

[4]<\/a> See supra <\/em>note 2, at \u00a7 4(1).<\/p>\n\n\n\n

[5]<\/a> Despite the technical restraints that have long applied to NPCs, their dialog can be highly entertaining.  See, e.g.<\/em>,  Lauren Quigley, A Day in Skyrim: An Analysis of NPCs & Storytelling in Video Games<\/em><\/a>, Medium (Feb. 20, 2015).<\/p>\n\n\n\n

[6]<\/a> The process of integrating AI chatbots into NPC functionality has not been quick or easy and developers are still grappling with the trade-offs.  The main friction points in current builds include latency (a delay while the AI generates a line), hallucinations (NPCs breaking character or breaking the lore), and the computational cost of running massive text-prediction engines concurrently for multiple characters.  There is also a concern about possible player backlash.  See <\/em>John Herrman, Gamers Are Extremely Mad About AI<\/em><\/a>, New York Magazine (Dec. 18, 2025).<\/p>\n\n\n\n

[7]<\/a> A leading example of this is inZOI<\/em><\/a>, a life simulation game developed and published by Krafton, the South Korean gaming giant.  The game integrates generative AI through its \u201cSmart Zoi<\/a>\u201d feature, allowing open-world NPCs to possess persistent memory and unscripted dialogue driven by on-device language models. See<\/em> NVIDIA GeForce NVIDIA ACE | inZOI – Create Simulated Cities with Co-Playable Characters<\/em><\/a>, YouTube (Jan. 7, 2025).  Rather than relying on massive, cloud-dependent Large Language Models (LLMs) that introduce response latency issues, Krafton addresses the hardware constraints of real-time gaming by utilizing lightweight, game-specialized Small Language Models (sLMs). See<\/em> Krafton AI<\/a> (last accessed May 29, 2026).  Krafton\u2019s stated goal<\/a> \u201cis to develop game-specialized language model agents capable of comprehension, reasoning, decision-making, and emotional expression.\u201d Id.  <\/em>Many of the AI implementations in games are still at an early stage of development.  Ubisoft<\/a>, the publisher behind major franchises such as Assassin\u2019s Creed<\/em>, Far Cry<\/em>, and Rainbow Six Siege<\/em>, has developed Ubisoft Teammates, \u201can experimental AI gaming project\u201d that enables players to interact with AI-powered squadmates through natural voice commands, allowing NPCs to understand context, assist with gameplay, and respond dynamically in real time.  Ubisoft Reveals Teammates-An AI Experiment to Change the Game<\/em><\/a>, Ubisoft News (Nov. 21, 2025) (last accessed June 1, 2026).  Chinese game publisher HoYoverse, best known for the blockbuster Genshin Impact<\/em>, reportedly plans to invest up to $14.6 billion in an in-house AI ecosystem over the next three years to support game development tools, with a particular focus on AI-powered NPCs, automation, content generation, and live-service operations.  Sophie McEvoy, HoYoverse to Invest up to $14.6bn in AI for In-House Tools<\/em><\/a>, GamesIndustry.biz (May 26, 2026) (last accessed June 1, 2026). <\/p>\n\n\n\n

[8]<\/a> Companies such as Epic Games, Unity, and NVIDIA are developing increasingly sophisticated tools that allow game developers to create hyper-realistic digital humans capable of both cinematic visual fidelity and AI-driven interaction. Through Unreal Engine<\/a>, Epic has positioned realistic human creation as an important part of modern game development, while NVIDIA is building a suite of technologies<\/a> \u201cdesigned to help developers bring digital humans to life for gaming use cases.\u201d  Together, these technologies are laying the groundwork for a new generation of AI-powered NPCs that behave more like interactive virtual actors than traditional scripted game characters.  See<\/em> Georgy Dev, OpenAI Integration in Unreal Engine – AI Chat & Text-to-Speech Tutorial<\/em><\/a>, Unreal Engine Dev Community (last updated Apr. 17, 2026); AI Integration in Unity: Enhancing NPC Behavior<\/em><\/a>, Appwill (June 19, 2025); How AI Could Disrupt Video-Gaming<\/em><\/a>, The Economist (Apr. 5, 2023).<\/p>\n\n\n\n

[9]<\/a> Replika is advertised as \u201ca friend to do life with.\u201d Replika website<\/a> (last accessed May 29, 2026).  Character.AI is a conversational platform where users can create, customize, and roleplay with millions of distinct, community-generated personas ranging from fictional characters to historical figures.  Character.AI website<\/a> (last accessed May 29, 2026).<\/p>\n\n\n\n

[10]<\/a> The exact count depends on how broadly one defines \u201cAI companion\u201d or \u201cchatbot\u201d legislation.  The following states have enacted laws specifically governing \u201ccompanion\u201d AI chatbots:  Cal. Bus. & Prof. Code \u00a7\u00a7 22601\u201322602 (West 2025), available at California SB 243<\/a>; Colo. H.B. 26-1263, 75th Gen. Assemb., 2d Reg. Sess. (Colo. 2026), available at Colorado HB 6-1263<\/a>; Conn. S.B. 5, 2026 Gen. Assemb., Reg. Sess. (Conn. 2026), available at Connecticut SB 5<\/a>; Ga. S.B. 540, 2025\u20132026 Reg. Sess. (Ga. 2026), available at Georgia SB 540<\/a>; Haw. S.B. 3001, 33d Leg., Reg. Sess. (Haw. 2026), available at Hawaii SB 3001<\/a> (the legislature has passed a bill that is awaiting the Governor\u2019s signature); Idaho S.B. 1297, 68th Leg., 2d Reg. Sess. (Idaho 2026), available at Idaho SB 1297<\/a>; Iowa Code \u00a7\u00a7 554J.1\u2013.6 (2027); Neb. L.B. 525, 109th Leg., 1st Sess. (Neb. 2025), available at Nebraska LB 525<\/a>; N.H. Rev. Stat. Ann. \u00a7 507:8-k (2025), available at New Hampshire Revised Statutes<\/a>; N.Y. Gen. Bus. Law \u00a7\u00a7 1700\u20131704 (McKinney 2025), available at New York General Business Law Article 47<\/a>; Or. S.B. 1546, 83d Legis. Assemb., Reg. Sess. (Or. 2026), available at Oregon SB 1546<\/a>; R.I. H. 7350, 2026 Gen. Assemb., Jan. Sess. (R.I. 2026), available at Rhode Island H 7350<\/a>; Wash. H.B. 2225, 69th Leg., Reg. Sess. (Wash. 2026), available at Washington HB 2225<\/a>.<\/p>\n\n\n\n

[11]<\/a> See supra<\/em> note 2.<\/p>\n\n\n\n

[12]<\/a> See supra <\/em>note 2, at \u00a7 1, 7, 15, 18, 33.<\/p>\n\n\n\n

[13]<\/a> See supra <\/em>note 2, at \u00a7 4.<\/p>\n\n\n\n

[14]<\/a> Conn. Off. of the Governor, Governor Lamont Signs Legislation Establishing Youth Online Safety Protections, Regulations Over Artificial Intelligence, and Initiatives to Upskill Connecticut\u2019s Workforce<\/em><\/a> (June 2, 2026); see also<\/em> Rachel Metz & Ellen Huet, Can a Chatbot Be Held Responsible for a Death<\/em><\/a>?<\/em>, Bloomberg (July 28, 2026).<\/p>\n\n\n\n

[15]<\/a> See supra <\/em>note 2, at \u00a7 5.<\/p>\n\n\n\n

[16]<\/a> See supra <\/em>note 2, at \u00a7 5(A)(i).<\/p>\n\n\n\n

[17]<\/a> See supra <\/em>note 2, at \u00a7 5(B)(i).<\/p>\n\n\n\n

[18]<\/a> See supra <\/em>note 2, at \u00a7 5(B)(ii)(2)(b).<\/p>\n\n\n\n

[19]<\/a> <\/sup>See<\/em> S.B. 1295, 2025 Gen. Assemb., Reg. Sess. (Conn. 2025), available at Connecticut SB 1295<\/a>; S.B. 2420, 89th Leg., R.S. (Tex. 2025), available at Texas SB 2420<\/a>; Cal. Civ. Code \u00a7\u00a7 1798.99.28\u2013.40 (West 2023)<\/a>.<\/p>\n\n\n\n

[20]<\/a> Exec. Order No. 14365<\/a>, 90 Fed. Reg. 58,499 (Dec. 11, 2025).<\/p>\n\n\n\n

[21]<\/a> See <\/em>Dustin Wong, et al., When Fantasy Feels Too Real: The Psychological Toll of Deep Media Attachment<\/em><\/a>, 17 Cureus 1, 1 (Aug. 16, 2025) (\u201cSocial media platforms (e.g., Twitter, TikTok, and Instagram) and entertainment platforms (e.g., YouTube, Kick, and Twitch) have provided users with unprecedented access to media figures, enabling individuals to foster a sense of intimacy and connection.\u201d).<\/p>\n\n\n\n

[22]<\/a> The term was coined by psychologists Donald Horton and R. Richard Wohl in 1956 just as television was becoming widely available.  They noted that some TV viewers began to develop the \u201cillusion of intimacy\u201d with the characters on the screen.  Parasocial Relationships<\/em><\/a>, Psychology Today (July 1, 2026).<\/p>\n\n\n\n

[23]<\/a> See <\/em>Andi H. Muhmin, et al., The Dynamics of Parasocial Relationships in the Age of Social Media: A Systematic Review and Future Research Agenda<\/a><\/em>, 9 J. Bus. & Behav. Entrepreneurship 1, 2 (June 2, 2025) (explaining that social media has changed how individuals digitally interact with others, causing an increase in parasocial relationships).<\/p>\n\n\n\n

[24]<\/a> <\/sup>See <\/em>Ailyn D. Diaz, et al., Clinical Assessment and Implications of Parasocial Relationships in Adolescents<\/em><\/a>, 65 J. Am. Acad. of Child & Adolescent Psychiatry 331, 331 (Mar. 2026)  (\u201cAdolescents are particularly vulnerable to PSRs [parasocial relationships] as they begin to separate from their primary caregivers and seek independence.\u201d); Tracy R. Gleason, et al., Parasocial Interactions and Relationships in Early Adolescence<\/a><\/em>, 8 Frontiers Psych. 1, 1 (Feb. 22, 2017) (explaining that adolescents show greater interest in popular media figures than other age groups, and their associated parasocial relationships \u201ctend to be intense\u201d).<\/p>\n\n\n\n

[25]<\/a> The term \u201cindustry standards\u201d is used in the statute without definition.  See supra <\/em>note 2, at \u00a7 6(a)(1).<\/p>\n\n\n\n

[26]<\/a> Id.<\/em> \u00a7 6(a)(1).<\/p>\n\n\n\n

[27]<\/a> Id.<\/em> \u00a7 6(a)(1)(F).<\/p>\n\n\n\n

[28]<\/a> Id.<\/em> \u00a7 6 (b).<\/p>\n\n\n\n

[29]<\/a> Id.<\/em> \u00a7 4(1).<\/p>\n\n\n\n

[30]<\/a> Id.<\/em><\/p>\n\n\n\n

[31]<\/a> See generally<\/em> N.Y. Gen. Bus. Law \u00a7\u00a7 1700-1704 (McKinney 2025); Neb. L.B. 525, \u00a7 2(1), 109th Leg., 1st Sess. (2025).<\/p>\n\n\n\n

[32]<\/a> Alex Weprin, Disney, Epic Games Using AI to Bring James Earl Jones\u2019 Darth Vader Voice to \u2018Fortnite\u2019<\/em><\/a>, Hollywood Reporter (May 16, 2025).<\/p>\n\n\n\n

[33]<\/a> Dani Di Placido, Fortnite\u2019s AI-Generated Darth Vader Controversy, Explained<\/em><\/a>, Forbes (May 20, 2025).<\/p>\n\n\n\n

[34]<\/a> See <\/em>Cal. S.B. 243, \u00a7 22601(b)(2)(B), 2025-2026 Reg. Sess. (2025); Idaho S.B. 1297, \u00a7 48-2102(2)(b)(ii), 68th Leg., 2d Reg. Sess. (2026); Or. S.B. 1546, \u00a7 1(1)(b)(ii), 83d Legis. Assemb., Reg. Sess. (2026); Wash. H.B. 2225, \u00a7 2(1)(b)(ii), 69th Leg., Reg. Sess. (2026).<\/p>\n\n\n\n

[35]<\/a> See<\/em> That Dragon, Cancer<\/em> (Numinous Games Jan. 11, 2016), available on Steam<\/a> (last accessed May 29, 2026).  The game is an interactive retelling of a 4-year-old\u2019s fight against cancer.  In the game description, the developers explain \u201cThis is where we go to remember our son Joel . . . We want to show you who he was, and how his life changed us.  Can we walk here together for a while?\u201d Id.<\/em>  One review of the game commented: \u201cI think [the game] might have changed me in some way that I\u2019m yet to fully understand . . .  If you play this game, it may change you.\u201d Polygon review quoted on Steam.  Id.<\/em>; see also<\/em> OMORI<\/em> (OMOCAT LLC 2020), available on Steam<\/a> (last accessed May 29, 2026).  The game portrays \u201cconcepts such as anxiety<\/a>, depression<\/a>, psychological trauma<\/a>, self-harm<\/a>, and suicide<\/a>.\u201d  Id<\/em>. The reviews on Steam are \u201coverwhelmingly positive.\u201d  Id<\/em>.  OMORI<\/em> was nominated for several awards, winning DreamHack\u2019s<\/a> \u201cDaringly Dramatic\u201d category in 2021, presented \u201cto the game that caused the biggest reaction and discussion in the community.\u201d Id.<\/em><\/p>\n\n\n\n

[36]<\/a> See<\/em> Cry of Fear<\/em> (Team Psykskallar Apr. 25, 2013), available on Steam<\/a> (last accessed May 29, 2026).  From the description of the game available on Steam: \u201cYou play as a young man desperately searching for answers in the cold Scandinavian night, finding his way through the city as he slowly descends into madness.\u201d Id.<\/em><\/p>\n\n\n\n

[37]<\/a> See<\/em> Pentiment<\/em> (Obsidian Entertainment 2022), available on Steam<\/a> (last accessed May 29, 2026).  The game contains \u201cpublic executions, stylized gore, suicide, references to abortion, frequent strong language, references to sexual assault, and references to sexual activities.\u201d  Id.<\/em>  Pentiment<\/em> won a Peabody Award in 2023.  Peabody award profile<\/a> (last accessed May 29, 2026).  The award was given for the game\u2019s \u201cinnovative fusion of sweeping historical sagas with smaller personal tales, and for exemplifying player agency in storytelling.\u201d  Id.<\/em><\/p>\n\n\n\n

[38]<\/a> See<\/em> Ollie Toms, The Witcher 3 romance: All romance options and endings explained<\/em><\/a>, Rock Paper Shotgun (Dec. 16, 2022).  The Witcher 3: Wild Hunt<\/em> was a massive commercial and critical success upon release in 2015, selling more than 6 million copies in its first six weeks and ultimately winning over 250 awards, including two trophies at The Game Awards in 2015.  See<\/em> CD Projekt website<\/a> (last accessed May 29, 2026). <\/p>\n\n\n\n

[39]<\/a> See supra<\/em> note 2, at \u00a7 1(c).<\/p>\n\n\n\n

[40]<\/a> Id.<\/em> \u00a7 5(A).<\/p>\n\n\n\n

[41]<\/a> Id.<\/em> \u00a7 33.<\/p>\n\n\n\n

[42]<\/a> Id.<\/em> \u00a7 33(b)-(d).<\/p>\n\n\n\n

[43]<\/a> Id.<\/em> \u00a7 33(e).<\/p>\n\n\n\n

[44]<\/a> Id.<\/em><\/p>\n\n\n\n

[45]<\/a> Id.<\/em><\/p>\n\n\n\n

[46]<\/a> See supra <\/em>note 2, at \u00a7 4(1).<\/p>\n","protected":false},"excerpt":{"rendered":"

AI Companion Legislation in the United States:What It Means for the Video Game Industry By: Stuart Irvin, Monique N. Bhargava, and Kalid Qur’an Knox[1] Across the United States, state legislatures are moving aggressively to regulate AI-powered chatbots and conversational companions, particularly those that interact with minors in a manner that provides human-like social interactions.  Driven […]<\/p>\n","protected":false},"author":254,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"disabled","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":true,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_post_was_ever_published":false,"_members_access_role":[],"_members_access_error":""},"categories":[28,1],"tags":[],"ppma_author":[399,401,400],"class_list":["post-4215","post","type-post","status-publish","format-standard","hentry","category-commentary","category-updates"],"jetpack_shortlink":"https:\/\/wp.me\/peZjrR-15Z","jetpack_sharing_enabled":true,"authors":[{"term_id":399,"user_id":254,"is_guest":0,"slug":"stuart-irvin","display_name":"Stuart Irvin","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/13a8a23a4a888837b9ee0a681a79e1bdd3517fa0a556759c028bc514aa9f8d63?s=96&d=mm&r=g","author_category":"","first_name":"","last_name":"","user_url":"","job_title":"","description":""},{"term_id":401,"user_id":256,"is_guest":0,"slug":"monique-n-bhargava","display_name":"Monique N. Bhargava","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/af16a0bd1c515cd5e8ffd6d5c066037205d7ed0c654ead2fcb8ab63ec82978aa?s=96&d=mm&r=g","author_category":"","first_name":"","last_name":"","user_url":"","job_title":"","description":""},{"term_id":400,"user_id":255,"is_guest":0,"slug":"kalid-q-knox","display_name":"Kalid Q. Knox","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/90da094756bd0b670d32b68e1bc7bf4dde021d1bd495dd5d10e24643a7c9482c?s=96&d=mm&r=g","author_category":"","first_name":"","last_name":"","user_url":"","job_title":"","description":""}],"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/posts\/4215","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/users\/254"}],"replies":[{"embeddable":true,"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/comments?post=4215"}],"version-history":[{"count":7,"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/posts\/4215\/revisions"}],"predecessor-version":[{"id":4225,"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/posts\/4215\/revisions\/4225"}],"wp:attachment":[{"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/media?parent=4215"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/categories?post=4215"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/tags?post=4215"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/journals.law.harvard.edu\/jsel\/wp-json\/wp\/v2\/ppma_author?post=4215"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}