Articles by tag: llm-security

AI zero-day: Google discovers machine-created exploit

Google confirms AI-created zero-day in criminal scheme. Learn how the vulnerability bypassed 2FA and what it means for your security. Read the report.

Attacks on LLM in Business: Vulnerabilities of Local Models

How Attackers Extract CRM Data Through LLM? Analysis of Real Attacks on Local and Cloud Models. Protection of SMB Solutions.

AI Infrastructure Wrapper: 2026 Trends for Developers

How stateful agents are changing the approach to AI development. Methods for controlling non-deterministic systems and safe integration into enterprise. Learn first.

LLM Vulnerability: Bypass via assistant prefill

Learn how sockpuppeting bypasses protections of 11 language models with one line. Risk analysis, protection measures, and business implications. Read attack details and recommendations.

Marusya and Salut Vulnerabilities: Bypassing Filters

Analysis of Technical Loopholes in Voice Assistants Marusya and Salut. Prompt injection via Choices, Memory, and Reminders. Protection Recommendations for LLM Developers.

PII Protection in LLM according to 152-FZ without UX loss

Implementation of PII token substitution for LLM on NestJS: compliance with 152-FZ, WebSocket streaming, PostgreSQL encryption. Instructions for developers of AI solutions in RF. Read details.