Safety Drift After Fine-Tuning: Evidence from High-Stakes Domains

TL;DR

This study investigates how safety properties of foundation models change after fine-tuning, revealing that safety is often inconsistent and unpredictable across different benchmarks and domains.

Contribution

It provides empirical evidence that safety assessments on base models do not reliably predict safety after fine-tuning, highlighting the need for re-evaluation in deployment contexts.

Findings

Fine-tuning causes large, heterogeneous safety changes.

Models often improve safety on some tests but degrade on others.

Safety behavior is not stable under downstream adaptation.

Abstract

Foundation models are routinely fine-tuned for use in particular domains, yet safety assessments are typically conducted only on base models, implicitly assuming that safety properties persist through downstream adaptation. We test this assumption by analyzing the safety behavior of 100 models, including widely deployed fine-tunes in the medical and legal domains as well as controlled adaptations of open foundation models alongside their bases. Across general-purpose and domain-specific safety benchmarks, we find that benign fine-tuning induces large, heterogeneous, and often contradictory changes in measured safety: models frequently improve on some instruments while degrading on others, with substantial disagreement across evaluations. These results show that safety behavior is not stable under ordinary downstream adaptation, raising critical questions about governance and deployment practices centered on base-model evaluations. Without explicit re-evaluation of fine-tuned models in deployment-relevant contexts, such approaches fall short of adequately managing downstream risk, overlooking practical sources of harm -- failures that are especially consequential in high-stakes settings and challenge current accountability paradigms.