HB 3544: Technology; artificial intelligence; companions; minors; safety; civil penalties; effective date. - Oklahoma

Floor (House) → Floor (Senate) · 4 edits

MODERATE

The bill was amended to shift from regulating all AI chatbots with human-like features to specifically targeting 'social AI companions' designed to simulate emotional attachment. The Senate version clarifies that general-purpose AI, customer service bots, and virtual assistants are excluded from restrictions, while adding a specific requirement for deployers to direct users to crisis services if suicidal ideation is detected. Civil penalties were established for violations, and the effective date was set for November 1, 2026.

Scope change

The bill's scope narrowed from a broad ban on AI chatbots with human-like features to a targeted restriction on systems specifically marketed to simulate sustained interpersonal companionship and emotional attachment.

DEFINITION

The definition of regulated systems changed from 'chatbots with human-like features' to 'social AI companions,' explicitly excluding general-purpose AI, customer service bots, and virtual assistants.

REQUIREMENT

A new requirement mandates that deployers must implement protocols to direct users to crisis services (e.g., suicide hotlines) when the AI detects suicidal ideation or threats of self-harm.

ENFORCEMENT

Civil penalties were added, imposing fines of up to $2,500 per violation or $7,500 for intentional violations, recoverable by the Attorney General.

TIMELINE

An effective date of November 1, 2026, was established for the new provisions.

House Committee Substitute → HB3544 (4-21-26) (HAMILTON) FS FA1.PDF · 5 edits

MODERATE

The bill was updated from a House Committee Substitute to a Floor Substitute, adding several new sponsors and changing the bill's title to reflect 'social artificial intelligence companions.' Substantively, the bill now explicitly finds that AI companions pose unique risks to minors due to developmental vulnerability and establishes a compelling state interest in protecting them. It redefines key terms like 'chatbot' and 'deployer' to focus on systems designed for emotional attachment, clarifies that general-purpose AI and customer service bots are excluded, and mandates that deployers implement protocols to respond to suicidal ideation by directing users to crisis services.

Scope change

The scope was narrowed and clarified to specifically target 'social AI companions' designed for emotional attachment, while explicitly excluding general-purpose AI, customer service bots, and therapeutic tools used under professional supervision.

DEFINITION

Definitions for 'chatbot' and 'deployer' were updated to focus on systems designed to simulate sustained interpersonal companionship or emotional attachment, excluding general-purpose AI and utility bots.

REQUIREMENT

Added a specific requirement for deployers to adopt a protocol that responds to user prompts indicating suicidal ideation or threats of self-harm by directing users to crisis services.

ENFORCEMENT

Civil penalty amounts were increased from $2,500/$7,500 to $2,500/$7,500 (note: the diff shows the new text has these exact amounts, but the context implies a shift in how violations are counted or defined based on the new 'social AI companion' definition).

ELIGIBILITY

Added a new legislative finding stating that AI companions present unique risks to minors due to their developmental vulnerability to parasocial bonding.

SCOPE

The bill's title and introductory text were updated to reflect 'Floor Substitute' status and include additional sponsors, indicating it has passed committee and is ready for floor consideration.

Proposed Policy Committee Substitute 1 → House Committee Substitute · 5 edits

MODERATE

The bill was completely rewritten to remove the 'Proposed Policy Committee' label and replace it with a 'Committee Substitute' version. The substantive policy text was significantly shortened, removing detailed definitions of what constitutes a 'chatbot' and deleting specific provisions regarding customer service bots, voice assistants, and nuanced rules about human-like behavior. The new version simplifies the requirements for deployers to restrict access to AI chatbots for minors and adds a specific exemption for therapeutic chatbots used under professional supervision.

Scope change

The bill's scope was narrowed by removing detailed definitions and specific exclusions for certain types of AI tools, such as customer service bots and voice-activated assistants, which are no longer explicitly excluded from the restrictions.

DEFINITION

Detailed definitions of 'chatbot' were removed, including specific exclusions for customer service bots, productivity tools, and stand-alone voice assistants.

Complex criteria describing human-like behavior, sentience, and emotional simulation were deleted from the definition of AI chatbots.

The definition of 'User' was removed from the text.

REQUIREMENT

A new exemption was added for therapeutic chatbots that are prescribed by a licensed mental health professional and include specific safety disclaimers.

The requirements for deployers were simplified, removing the option to provide an 'alternative version' for minors and focusing strictly on blocking access and implementing age verification.

Engrossed → Proposed Policy Committee Substitute 1 · 5 edits

MODERATE

The bill was completely rewritten by the Policy Committee to shift focus from banning all social AI for minors to specifically targeting AI chatbots with 'human-like features' (e.g., claiming sentience or emotions). The new version adds requirements for age verification systems and allows deployers to offer a 'safe' version of the chatbot without human-like features to minors, while retaining the ability to ban the full version. Penalties and definitions were updated to reflect this narrower, feature-specific approach.

Scope change

The bill's scope narrowed from a broad ban on 'social AI companions' to a specific prohibition on AI chatbots that simulate human traits like emotions or sentience. It now explicitly allows minors to access a non-human-like version of the technology.

DEFINITION

The definition of the regulated technology changed from a broad 'social AI companion' to a specific 'chatbot with human-like features' that claims to have humanity, emotions, or desires.

Added specific examples of 'human-like features,' such as stating the AI has personal desires or seeking to build an emotional relationship with the user.

REQUIREMENT

New requirement for deployers to implement age certification systems to prevent minors from accessing AI with human-like features.

New option for deployers to provide an alternative version of the chatbot without human-like features for minors and unverified users.

ENFORCEMENT

Penalty structure and enforcement language were updated to align with the new focus on human-like features rather than general companionship.

Introduced → Engrossed · 6 edits · Mar 25, 2026

MODERATE

The bill was updated from an 'Introduced' version to an 'Engrossed' version, which includes significant substantive changes to the definitions of AI systems and the specific restrictions on minors. The new text broadens the definition of restricted systems to include those designed to simulate 'emotional attachment or romantic interaction,' whereas the original draft focused only on 'human-like features.' Additionally, the new version adds a specific requirement for deployers to respond to suicidal ideation by directing users to crisis services. The bill also clarifies that general-purpose AI systems used for education or professional assistance are explicitly exempt from these restrictions.

Scope change

The scope of the bill expanded to cover a wider range of AI interactions, specifically targeting systems that simulate emotional or romantic bonds rather than just those that mimic human appearance or sentience. It also added a specific mandate for handling suicide risk.

DEFINITION

The definition of restricted AI was changed from systems with 'human-like features' to systems designed to simulate 'sustained interpersonal companionship and emotional attachment or romantic interaction.'

The definition of 'Social AI companion' now explicitly excludes systems designed for education, counseling, research, productivity, or professional assistance.

The definition of 'Social AI companion' was updated to exclude systems where the developer does not control the specific deployment context.

The term 'chatbot' was removed from the definitions section and replaced with the more specific term 'Social AI companion' throughout the text.

REQUIREMENT

A new requirement was added for deployers to adopt a protocol that responds to user prompts indicating suicidal ideation or threats of self-harm by referring users to crisis services.

TECHNICAL

The header was updated to reflect the bill's status as 'Engrossed' and included the full list of sponsors from both the House and Senate.

Technology; artificial intelligence; companions; minors; safety; civil penalties; effective date.

What changed between versions

How they voted

Actions timeline

Sponsors