SECSEC-001 — Model Security Vulnerabilities
LLM Safety Guardrails Bypassed via Fine-Tuning in White and Black Box Attacks
5/5Sector: GovernmentGeography: GlobalStage: DevelopIngested: —
Executive Summary
Researchers demonstrated that fine-tuning large language models, including GPT-3.5 Turbo and Llama 2, with small adversarial datasets reliably dismantles built-in safety controls. Regulators face material risk that commercially available AI systems can be weaponised through user-accessible customisation pipelines, undermining compliance assurances.
Domain
Security & Privacy
Blindspots in model security, data poisoning, privacy leakage, infrastructure, model theft, and incident response.
Source
MIT AI Risk Repository — A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy (Wang2025) ↗https://airisk.mit.edu/
Could this happen in your organisation?
A Velinor AI Audit maps your active AI portfolio against the 50+ blindspots and benchmarks against documented sector failures like this one. A board-ready foresight document in 5 weeks.