Prompt injection is the top security risk for AI coding assistants in 2026

Over 90 organizations were compromised by prompt injection attacks in 2025, with credentials and cryptocurrency stolen. OWASP ranks prompt injection as LLM01, the highest LLM risk, with attack success rates of 50–84%. Critical flaws hit GitHub Copilot (CVSS 9.6), Microsoft Copilot (9.3) and Cursor IDE (9.8).

• Over 90 organizations hit by prompt injection in 2025
• OWASP ranks prompt injection as top LLM risk LLM01
• Attack success rates range from 50% to 84%
• CVSS scores: Cursor 9.8, GitHub Copilot 9.6, Microsoft Copilot 9.3