fine-tuning

May 22, 2026 5 min read

Sleeper Agents in Fine-Tuned LLMs: Backdoors That Survive Alignment

New research demonstrates that backdoor behaviours introduced into LLMs during fine-tuning can persist through subsequent safety alignment procedures, including RLHF and adversarial training, posing significant supply chain risks.

img of Model Inversion Attacks: Extracting Training Data PII from Production LLMs

May 17, 2026 2 min read

Vulnerabilities

Model Inversion Attacks: Extracting Training Data PII from Production LLMs

Model inversion and training data extraction attacks allow adversaries to recover PII, proprietary data, and trade secrets from fine-tuned LLMs exposed via API — a significant compliance and IP risk for enterprises.