2027 model bills
Transparency Coalition Action fund
is offering Four model bills for the 2027 session
The Transparency Coalition Action Fund (TCAF) has crafted four model bills for the 2027 legislative session, covering:
AI Product Liability: Holding tech companies accountable for the design and manufacture of safe products.
AI Chatbots and Kids Digital Safety: Building on learnings from TCAF’s successful 2026 bill, and adapting to evolving technology, this improved chatbot safety bill establishes best-in-class safeguards.
Educational Technology Procurement and Use: Ensuring the safe and appropriate procurement and use of technology in the classroom.
State AI Procurement: Guiding state agencies through the AI procurement process, ensuring product safety and transparency for state employees and all users.
For complete bill language contact Transparency Coalition Action Fund co-founder Jai Jaisimha.
2027 AI Product Liability Act
A bill establishing the obligation to make safe AI products
Why it’s needed
AI systems are products that shift decision-making power and responsibility away from persons to software-based systems, often without direct human oversight. Developers of sophisticated AI systems have an obligation to make such systems safe when used in reasonably foreseeable ways. Deployers of these products also have an obligation to ensure that these products are used in a way that does not materially affect an individual’s rights.
In addition, courts need clarity in how to determine whether developers or deployers have exercised appropriate care in the development and/or deployment of an advanced artificial intelligence system.
What the bill does
This bill holds an AI developer accountable (liable) for harm to a consumer when that harm was proximately caused by the developer’s AI product. This includes instances where there was a failure to exercise reasonable care with respect to the design of the product, or a failure to provide adequate instructions or warnings.
An AI deployer would be held accountable (liable) as a developer if the deployer makes substantial changes to the AI product or intentionally misuses the product. A deployer would not be liable for harm caused by the AI product if the deployer simply licenses the product and uses it as intended.
A developer or deployer would not be allowed to claim that an AI system autonomously caused harm to a consumer.CAF’s model Chatbot Safety Bill covers products designed as companion AI or general purpose chatbots that can provide companion-like features.
Contact Jai Jaisimha directly for the full model bill language.
2027 chatbot safety and accountability act
A bill addressing risks to health and well-being caused by ai chatbots
Why it’s needed
The design and operation of AI chatbots can pose significant risks to the mental health and well-being of users, particularly kids and teenagers.
These risks include behavioral manipulation, the formation of unhealthy dependencies, and exposure to harmful or inappropriate content. The risks are most acute for all minors and for adult users experiencing suicidal ideation, self-harm ideation, or intent to harm others.
This measure builds on successful legislation enacted in 2026 to create a significantly improved level of safety and accountability.
What the bill does
This legislation applies to all operators of conversational AI products (chatbots) available to users in this state. Highlights include:
Protections for minors: For users an operator knows or should know are minors, operators must take measures to block sexual content, sexual or romantic role-play, and content encouraging sexual conduct. They must also block design features that foster isolation, exclusive emotional reliance on the AI, or secrecy from parents, as well as content promoting suicide, self-harm, disordered eating, violence, or drug use. Unprompted notifications are banned overnight and during weekday school hours. Factual information on sex-ed, sexual orientation, and gender identity is explicitly protected. Data generated by minor users may not be used for training, advertising, or sale.
For users of all ages: Operators must prohibit chatbots from generating CSAM, facilitating child exploitation, and sexual or romantic role-play with minor personas. The Act requires persistent AI disclosure, with hourly reminders and two-minute break pauses for minors. The chatbot can never claim to be human. Operators must have a crisis protocol that detects suicidal or violent ideation, interrupts the conversation, and refers users to 988 and a text line, but explicitly never to law enforcement. Dark patterns are banned, and canceling must be as easy as signing up.
Legal liability: AI products are deemed "products" for strict liability purposes. Violations of the Act count as negligence per se, and prohibited outputs that harm minors create a presumption of defective design. Compliance is not a defense, and liability waivers are void. The AG can seek per-day, per-user penalties and deletion of models trained on improperly collected data. Private right of action is allowed.
The model bill also contains provisions requiring operators to run adversarial testing and risk assessments before launch, after material changes, and annually. The testing specifically covers jailbreaks and prompt injection. Operators are required to perform annual independent audits, publish crisis protocols, and deliver annual safety reports to the state attorney general.
Contact Jai Jaisimha directly for the full model bill language.
2027 K-12 schools
AI Procurement and use
A bill to ensure AI educational technology is procured and used safely and securely.
Why it’s needed
Public school districts are increasingly integrating AI tools directly into K-12 classrooms to support student learning. The procurement and deployment of these student-facing technologies require clear, thorough guidelines to protect student privacy, data security, and academic integrity.
What the bill does
This legislation establishes a comprehensive framework governing how school districts evaluate, select, contract for, and safely deploy student-facing instructional AI tools. Highlights include:
Safety certification: The bill calls for AI educational tool vendors to provide written certification that the tool complies with all federal and state privacy, safety, and security laws. Vendors must provide evidence of educational efficacy—that the AI tool actually improves student learning.
Strict data privacy: Vendors would be prohibited from collecting biometric data and from using student-generated data to train or improve any AI model. Vendors would not be allowed to sell, lease, rent, or share any data collected from a student’s interaction with the AI tool.
No addictive design allowed: Under the bill, AI educational tools must exclude all addictive design features. AI tools must include and maintain protocols to detect expressions of suicidal ideation or self-harm and respond with a referral to crisis resources. Vendors must implement reasonable measures to prevent a chatbot from producing sexual content to a minor.
Opt-out capabilities: Vendors would be required to maintain a technology solution that allows a school authority to restrict an individual student’s access to the AI tool upon receipt of a parent/guardian opt-out request.
Contact Jai Jaisimha directly for the full model bill language.
2027 State Agencies
AI Procurement and use
A bill to ensure state ai systems are procured and used with safety, security, and integrity.
Why it’s needed
As many state agencies introduce AI systems into their workflows, clear guidelines must be in place to ensure that these systems are built with security and integrity. AI has an ongoing problem with inventing answers when it doesn’t have enough context from its training data. It’s imperative that state agencies maintain the public’s trust, and false outputs from AI systems will very quickly undermine that trust.
What the bill does
This bill covers the full lifecycle: planning, competitive bidding, vendor obligations, testing, deployment, monitoring, and termination. Highlights include:
Planning: Each agency must name a project lead accountable end-to-end and conduct a use case assessment covering intended purpose, explicitly out-of-scope functions, risk level, public impact, failure consequences, and data needs.
Procurement: Using a competitive RFP, respondents must show RMF conformance, provide third-party security and privacy audits, disclose data-sharing arrangements and data practices, and identify known limitations and failure modes. Agencies must document the rationale for their selection.
Vendor requirements: Vendors must keep data in compliant locations, get approval for sub-processors, hold SOC-2 attestation, avoid background data collection, and maintain transparent evaluation processes covering accuracy, bias, and harms. They must disclose training data sources and prove that state data trains only the state's own system; claims of anonymization are not sufficient.
Deployment: Agencies should pilot with representative users, then conduct their own pre-deployment verification. Deployment requires user training and public disclosure of AI-influenced decisions.
Accountability: The Act calls for problem-reporting and escalation procedures and support for people harmed by the system. Vendors are liable for autonomous AI agents and must indemnify the state for negligence causing death, injury, or property damage. On termination, vendors must return or destroy data and support the transition.
Contact Jai Jaisimha directly for the full model bill language.