Meta SecAlign: Open-Weight Model for Prompt-Injection

Research question

Can a commercial-grade open model ship with built-in prompt-injection robustness?

Key findings

Llama-3.3-70B fine-tuned with SecAlign matches frontier models on utility while resisting most injections - Released openly to enable defence research

Sources

arxiv.orghttps://arxiv.org/abs/2507.02735

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit research Report a correction or request removal

Related in Research Paper Library

SecAlign

Trained models to prefer ignoring injected instructions using preference optimisation, followed by Meta SecAlign, an open foundation model with the defence built in.

Prompt injection defences by design2024

The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions

OpenAI trained models to treat system, developer, user and tool messages as descending levels of privilege, following lower-level instructions only when aligned with higher ones, which raised robustness to injection…

Prompt injection defences by design2024

The Instruction Hierarchy

OpenAI's instruction hierarchy training so that system, user and tool messages carry different privilege levels, deployed in GPT-4o mini and later models.

Prompt injection defences by design2024