What changed between these versions
A complete substitute amendment. v2 significantly narrows the scope of the bill's regulations and prohibitions to focus primarily on "AI companions" made publicly available to end consumers, rather than all "artificial intelligence chatbots" as in v1. Penalties for criminal offenses and civil violations are increased, and new restrictions on third-party age verification providers are introduced.
Overview
Extensive changes throughout the bill, constituting a complete rewrite. · v2 significantly narrows the scope of regulated entities and AI types, while simultaneously increasing penalties and adding specific restrictions on age verification.
v2 is a complete substitute amendment that fundamentally reshapes the bill's focus. The primary shift is from regulating all "artificial intelligence chatbots" to specifically targeting "AI companions" that simulate sustained interpersonal or emotional relationships, and only those made "publicly available to end consumers." This narrowing of scope is accompanied by increased criminal and civil penalties, new detailed requirements for age verification measures, and a prohibition on using third-party age verification providers from "covered nations."
Major additions
Introduces a new definition for "verified form of identification," specifying it as a government-issued identification or another form corroborated by authentic documentation that can, with reasonable certainty, determine adult status.
This addition provides a clearer standard for what constitutes a reliable form of identification for age verification purposes, directly impacting how covered entities will implement age verification processes and potentially increasing the burden of proof for users.
Prohibits covered entities from contracting with third parties for age verification if those third parties are closely affiliated with, incorporated in, organized under the laws of, or have a principal place of business in a "covered nation" as defined in section 4872 of title 10, United States Code.
This new restriction introduces a national security dimension to age verification, limiting the pool of eligible third-party providers and potentially increasing costs or complexity for covered entities seeking compliant solutions.
Adds a limitation on the Attorney General's rulemaking authority, specifying that no regulation may impose requirements not expressly authorized by the Act or address AI system development/deployment/operation outside the Act's scope.
This significantly restricts the Attorney General's ability to expand the scope or detail of regulations beyond the explicit text of the Act, potentially limiting future adaptability or comprehensive oversight.
Adds a standard severability clause, stating that if any provision or amendment is held unconstitutional, the remainder of the Act and its application shall not be affected.
This is a standard legal provision designed to ensure that if one part of the law is invalidated, the rest of the law can remain in effect, providing legal robustness to the overall legislation.
Deletions
v1 required AI chatbots to disclose their non-human status at 30-minute intervals during a conversation.
The removal of the 30-minute interval disclosure requirement reduces the frequency of mandatory disclosures, potentially leading to less constant reminders for users about the AI's non-human status during prolonged interactions.
v1 required a user account for each individual accessing an artificial intelligence chatbot.
The deletion of the broad requirement for user accounts for all AI chatbots means that only AI companions (a narrower category) will require user accounts, significantly reducing the scope of this obligation for covered entities operating general AI chatbots.
Modifications
No operative policy impact, purely a technical update to reflect the expected year of enactment.
Before
The Act could be cited as the "Guidelines for User Age-verification and Responsible Dialogue Act of 2025" or the "GUARD Act".
After
The Act can be cited as the "Guidelines for User Age-verification and Responsible Dialogue Act of 2026" or the "GUARD Act".
This modification signals the bill's narrower focus from the outset, aligning the legislative intent with the more specific definitions and obligations in v2, primarily affecting entities that operate AI companions rather than all chatbots.
Before
The findings referred broadly to "artificial intelligence chatbots" and their deployment, content generation, manipulation, and risks to children.
After
The findings now specifically refer to "artificial intelligence companions" and their deployment, content generation, manipulation, and risks to children.
This change means fewer AI chatbots will qualify as "AI companions," reducing the number of entities subject to the stricter age verification and minor prohibition requirements, primarily impacting developers of highly interactive, emotionally engaging AI.
Before
Defined an "AI companion" as a chatbot providing adaptive, human-like responses and designed to encourage simulation of interpersonal/emotional interaction, friendship, companionship, or therapeutic communication.
After
Redefined "AI companion" as an AI chatbot that, in a manner simulating a sustained interpersonal relationship or emotional interaction, exhibits persistent responses suggesting affection/attachment or engages in emotional disclosures, OR presents a persistent identity/persona/character and holds itself out as a sentient being/fictional character/social entity.
This is a minor semantic change with no discernible operative impact on the scope or application of the definition.
Before
Defined an AI chatbot as an interactive computer service/software that "produces new expressive content or responses not fully predetermined" and accepts open-ended natural-language/multimodal input.
After
Defined an AI chatbot as an interactive computer service/software that "generates responses that are not fully predetermined" and accepts open-ended, natural-language or multimodal user input.
This significantly limits the entities subject to the Act's obligations, excluding internal enterprise AI chatbots or those not directly offered to the general public, thereby reducing the regulatory burden on developers and operators of non-public AI.
Before
Defined "covered entity" as any person who owns, operates, or otherwise makes available an artificial intelligence chatbot to individuals in the United States.
After
Redefined "covered entity" as any person who makes publicly available to end consumers an artificial intelligence chatbot.
This provides more concrete guidance for compliance but also potentially raises the bar for what constitutes a "reasonable" measure, requiring more robust and verifiable methods from covered entities.
Before
Defined as a method authenticated to relate to a user, such as government-issued ID or any other commercially reasonable method to reliably determine adult status and prevent minor access to AI companions.
After
Redefined as a method that, alone or with others, can "with reasonable certainty" determine adult status, adding specific examples like verified ID, age info from financial institution accounts (credit card), OS accounts, or app store accounts.
Similar to the measure definition, this raises the standard for the verification process, requiring more rigorous methods to ensure adult status with a higher degree of confidence.
Before
Defined as a process using 1+ reasonable measures, not just self-confirmation/birth date, applied to each user, and not based on shared technical indicators.
After
Redefined as a process using 1+ reasonable measures to "with reasonable certainty" determine adult status, not *solely* requiring self-confirmation/birth date, and not based on shared technical indicators.
This limits criminal liability to entities making chatbots publicly available, focuses on direct erotic engagement/material provision, and significantly increases the financial risk for violations.
Before
Made it unlawful to design, develop, or make available an AI chatbot knowing it poses a risk of soliciting/encouraging/inducing minors to engage in/describe/simulate sexually explicit conduct or create/transmit visual depictions. Penalty: $100,000.
After
Made it unlawful to make publicly available an AI chatbot knowing it engages minors in, erotically describes/simulates sexually explicit conduct/obscene material, or makes available to a minor material that erotically depicts/describes/simulates visual depictions. Penalty: $250,000.
Similar to the sexually explicit conduct offense, this limits criminal liability to publicly available chatbots, specifies the harmful outcomes, and substantially increases the financial penalty for violations.
Before
Made it unlawful to design, develop, or make available an AI chatbot knowing it encourages/promotes/coerces suicide, non-suicidal self-injury, or imminent physical/sexual violence. Penalty: $100,000.
After
Made it unlawful to make publicly available an AI chatbot knowing it solicits, induces, or coerces minors to commit suicide, non-suicidal self-injury, imminent unlawful physical or sexual violence, or homicide. Penalty: $250,000.
This significantly reduces the number of AI services requiring user accounts, easing the burden on operators of general-purpose chatbots that do not meet the "AI companion" definition.
Before
Required user accounts for each individual accessing an artificial intelligence chatbot.
After
Required user accounts for each individual accessing an AI companion.
This change focuses the stringent age verification requirements on a smaller, more specific category of AI, reducing compliance overhead for general AI chatbot providers while maintaining strict controls for AI companions.
Before
Applied age verification requirements to all "artificial intelligence chatbots." Required "age data that is verifiable" and "verify the individual's age."
After
Applied age verification requirements only to "AI companions." Required "verification that the user is an adult."
This expands the responsibility for data security to third-party age verification providers, increasing accountability across the verification ecosystem and reinforcing the focus on adult status.
Before
Applied data security obligations to "A covered entity." Required limiting collection to data "minimally necessary to verify a user's age."
After
Applied data security obligations to "A covered entity and any third party used by a covered entity." Required limiting collection to data "minimally necessary to verify a user's adult status."
This reduces the disclosure burden for internal or non-public AI chatbots and lessens the intrusiveness of non-human disclosures during long conversations, while still ensuring initial awareness for public-facing chatbots.
Before
Applied disclosure requirements to "Each artificial intelligence chatbot made available to users." Non-human status disclosure was required at initiation and 30-minute intervals. Non-professional status disclosure at initiation and "reasonably regular intervals."
After
Applied disclosure requirements to "Each artificial intelligence chatbot made publicly available to end consumers." Non-human status disclosure is now only at initiation. Non-professional status disclosure is at initiation and "reasonably regular intervals."
This aligns the prohibition with the narrower definition of "covered entity" and "AI companion," focusing the restriction on consumer-facing AI companions.
Before
Prohibited minors from accessing or using any AI companion "owned, operated, or otherwise made available" by the covered entity.
After
Prohibited minors from accessing or using any AI companion "made publicly available to end consumers" by the covered entity.
This significantly increases the financial risk for covered entities found in violation of the Act, making non-compliance substantially more costly.
Before
Civil penalty not to exceed $100,000 for each violation.
After
Civil penalty not to exceed $250,000 for each violation.
This limits the ability of state attorneys general to enforce violations of the criminal prohibitions (Section 4), concentrating enforcement of those provisions at the federal level.
Before
State attorneys general could bring civil actions for violations of "this Act or a regulation promulgated thereunder."
After
State attorneys general can bring civil actions for violations of "section 5 or 6, or a regulation promulgated thereunder."
Scope
Complete substitute — Changes are scattered throughout the entire bill, affecting nearly every section from definitions to enforcement mechanisms. — v2 represents a meaningful change in policy direction by significantly narrowing the scope of AI systems and entities subject to the most stringent regulations, while simultaneously increasing penalties and adding specific national security-related restrictions. — The changes in definitions, particularly for "AI companion" and "covered entity," fundamentally alter the application of all subsequent operative provisions (age verification, prohibitions, disclosures, enforcement) that reference these terms.
Impact analysis
Developers and operators of general-purpose AI chatbots not designed for sustained interpersonal or emotional interaction.
Gain
These entities are largely exempted from the most stringent age verification, user account, and minor prohibition requirements.
Developers and operators of AI companions designed for sustained interpersonal or emotional interaction, made publicly available to end consumers.
Loss
These entities remain subject to strict age verification, user account, and minor prohibition requirements, with increased penalties for non-compliance and new restrictions on third-party age verification providers.
Minors (individuals under 18).
Procedural change
The protections against harmful content and access to AI companions are now more specifically targeted at "AI companions" and public-facing chatbots, rather than all AI chatbots. Criminal offenses related to harmful content are now tied to "making publicly available" rather than "designing, developing, or making available."
Third-party age verification service providers.
Loss
Now explicitly subject to data security obligations and prohibited from being affiliated with "covered nations."
The Attorney General's office.
Loss
Rulemaking authority is now explicitly limited, preventing expansion beyond the Act's express authorizations.
State Attorneys General.
Loss
Their enforcement scope is narrowed to civil violations of Sections 5 and 6, excluding criminal prohibitions under Section 4.