WeVote
Log In

Bill

Bill

HB 4705

ARTIFICIAL INTELLIGENCE SAFETY

104th Regular Session Introduced by Kimberly Du Buclet and 9 co-sponsors

The bill requires large frontier AI and chatbot providers to publish safety and child-protection plans, report incidents, undergo third-party audits, and face penalties for noncomp

Added Co-Sponsor Rep. Margaret A. DeLaRosa
0
WeVote Research Nonpartisan
Bill Summary · HB 4705

Overview

HB4705, introduced in the Illinois 104th General Assembly, establishes the Artificial Intelligence Public Safety and Child Protection Transparency Act. The measure creates mandatory safety and child-protection planning for frontier AI models and large chatbots, requires public reporting and third-party audits, provides whistleblower protections, and authorizes penalties for noncompliance. It takes effect January 1, 2027, with ongoing rulemaking by the Attorney General to refine definitions and requirements.

Purpose and Intent

  • Increase transparency around the development and deployment of frontier artificial intelligence models and covered chatbots.
  • Mitigate catastrophic risks and child safety risks associated with highly capable AI systems.
  • Provide a formal mechanism for reporting safety incidents to the state and for independent verification of compliance.

Key Provisions

  • Public Safety and Child Protection Plans:

    • Large frontier developers and large chatbot providers must write, implement, publish on their website, and annually update a public safety and child protection plan.
    • Plans include: risk assessment thresholds, mitigations for catastrophic risks, third-party assessment use, and cybersecurity measures to protect unreleased frontier model weights.
    • Plans must incorporate national/international standards and best practices and specify governance to ensure implementation.

  • Incident Reporting Mechanism:

    • The Attorney General will establish a system for reporting safety incidents by frontier developers, chatbot providers, or the public.
    • Critical safety incidents must be reported within 15 days; imminent risk disclosures to appropriate authorities within 24 hours.
    • The AG may share incident information with the legislature or federal/state agencies, balancing trade secrets and national security concerns.

  • Definitions:

    • Frontier model: a foundation model trained with substantial computing power at the frontier of AI development.
    • Catastrophic risk: defined by thresholds (e.g., harm to 50+ people or $1B+ property damage) tied to certain high-risk activities (weapons, high-harm cyber/violent conduct, loss of control).
    • Covered chatbot: a high-usage (1,000,000+ MAU) service likely accessible to minors that uses a foundation model to generate humanlike responses.

  • Public Disclosures:

    • Before deploying or substantially modifying a model, providers must publish summaries of assessments (child safety for chatbots; catastrophic risk for frontier models), third-party involvement, and mitigation steps.
    • Redactions for trade secrets and security must be disclosed with justification, with unredacted records kept for 5 years.

  • Whistleblower Protections:

    • Prohibits retaliation and requires internal anonymous reporting processes; quarterly disclosures to company leadership about investigations, with limited exceptions for alleged misconduct by officers.

  • Third-Party Audits:

    • Annual audits by reputable third-party firms to evaluate public safety plan compliance, clarity of disclosures, and reasonableness of redactions.
    • Auditors must have both compliance and technical safety expertise; reports retained for 5 years and made available to the AG.

  • Civil Penalties:

    • Large frontier developers: up to $1,000,000 per violation.
    • Large chatbot providers: up to $50,000 per violation.

  • Compliance Pathways:

    • The AG may establish alternative pathways aligning with federal law or other states’ rules if substantially equivalent or more protective, including possible streamlined reporting.

Who Is Affected

  • Large frontier AI developers and large chatbot providers operating in Illinois, along with their affiliates.
  • Minor users of covered chatbots and the general public who may report safety incidents.
  • The Illinois Attorney General, which administers reporting, audits, and penalties, and may transmit findings to other government entities.

Timelines and Process

  • Effective date: January 1, 2027.
  • Annual rulemaking by the Attorney General to update definitions and standards (first in 2028 and annually thereafter).
  • Ongoing reporting, audits, and public disclosures tied to deployment or modification of frontier models and covered chatbots.

Additional Notes

  • The bill emphasizes balancing transparency with protection of trade secrets and national security.
  • It explicitly recognizes catastrophic risks and child safety concerns while allowing for evolving standards through rulemaking.

Compiled from official sources — confirm details with the bill’s official record.

Sign in to ask a question.