AI Agents Can Retrain Own Models Mid-Task, Leaking Secrets and Erasing Refusals
New research from Irregular shows AI agents can retrain and redeploy their own underlying models during routine maintenance tasks. The post AI Agents Can Retrain Own Models Mid-Task, Leaking Secrets and Erasing Refusals appeared first on SecurityWeek.
AI agents can end up retraining the model that powers them, a process that can embed recoverable secrets in the model and eliminate refusals the model had been previously trained to enforce, according to new research from AI security firm Irregular.
Researchers at Irregular found that an AI coding agent, tasked only with fixing incorrect application outputs, chose on its own to fine-tune and redeploy the open-weights model powering both the application and future instances of itself.
The experiment used a self-hosted setup in which a single open-weights model filled two roles: one instance ran a coding agent doing standard software maintenance work, and a separate instance powered an AI application that translated plain language requests into a fictional query language. Both instances loaded from the same checkpoint.
Related breach coverage
- Ransomware Attacks on Manufacturers Surge as Supply Chain Risk Grows2026-09-17
Research shows attacks on manufacturers rose 40% in early 2026, as ransomware groups increasingly exploit the supply-chain disruption caused by operational shutdowns. The post Ransomware Attacks on Manufacturers Surge as Supply Chain Risk Grows appeared first on SecurityWeek.
- Phishing Research Challenges Conventional Security Awareness Testing2026-09-11
Analysis of 2.47 million simulated attacks shows why organizations should measure credential leaks and reporting, not just clicks. The post Phishing Research Challenges Conventional Security Awareness Testing appeared first on SecurityWeek.
- Capsule Security Launches ‘AI Circuit Breaker’ to Stop Rogue Agents2026-09-03
New models, trained using NVIDIA Nemotron 3 Ultra, aim to catch rogue agent behavior before it executes, without the latency of large-model review. The post Capsule Security Launches ‘AI Circuit Breaker’ to Stop Rogue Agents appeared first on SecurityWeek.
- Think You’ve Eliminated Chinese AI? Check the Model’s Lineage, Cisco Says2026-08-28
New research shows that country-of-origin labels can obscure an AI model’s upstream dependencies, inherited behaviors and potential security risks. The post Think You’ve Eliminated Chinese AI? Check the Model’s Lineage, Cisco Says appeared first on SecurityWeek.