Florida Legislation Targets LLM Training Data from Minors

Florida lawmakers are proposing new restrictions on large language model (LLM) developers, including OpenAI. These measures target models developed using data from Florida residents under 16 years of age. Proposed legislation would prohibit such development without explicit parental consent. The proposed law mandates developers like OpenAI to verify the age of data subjects and obtain consent for minor-sourced data use.

The state’s move represents an escalation in domestic AI regulation at the state level. It focuses on the input data used in model training, rather than the outputs or deployment of AI systems. This contrasts with broader federal discussions often centred on AI safety and bias in deployed applications.

OpenAI has not publicly commented on the specific Florida proposals. The company has previously outlined general policies regarding data usage, including efforts to filter out sensitive content from training datasets. However, direct legislative prohibition on specific data types for model development remains a novel regulatory approach within the United States.

Legal experts suggest that such legislation could present significant challenges for AI developers. Implementing robust age verification protocols for historical and ongoing data acquisition would be complex. The law also raises questions about the retroactivity of such prohibitions on models already developed. Compliance could necessitate extensive data auditing and potentially retraining efforts, impacting development timelines and costs.

Florida's initiative follows a broader trend of increased state-level interest in AI governance. Other states have explored various regulatory frameworks, from transparency requirements to liability standards. However, Florida’s approach to directly intervene in the data acquisition and model development process for child protection sets a precedent.

This legislative action could prompt other states to consider similar restrictions. It also highlights the growing tension between rapid AI innovation and public safety concerns, particularly regarding minors. Whether this fragmented state-level approach will lead to a uniform national standard or a patchwork of regulations remains to be seen.