Definition
A 2026 Connecticut state law — one of the most comprehensive US state AI statutes to date — that places direct legal obligations on developers of the most powerful 'frontier' AI models, not just on companies that deploy AI products. Key provisions include a regulatory sandbox allowing companies to test new AI approaches under regulatory supervision, and specific safety requirements for AI products accessible to minors. Unlike most US state AI laws that focus on end-users of AI, SB 5 reaches upstream to the companies building the foundation models themselves.
Why it matters
Any company developing a large-scale AI model whose services touch Connecticut residents faces new disclosure, safety, and sandbox obligations — setting a precedent that frontier model liability can attach at the development stage, before any product ships.