Prompt Injection Remains Unsolved in Generative AI, OWASP Researcher Warns (infosecurity-magazine.com)

• Prompt injection remains architecturally unsolved because LLMs lack privilege separation between inputs.
• Agentic AI increases risk: injection can trigger real-world actions like data exfiltration or system compromise.
• Fogel calls for machine-speed monitoring and real-time containment rather than prevention only.

"At Infosecurity Europe 2026, OWASP contributor Ariel Fogel warned that prompt injection remains an unresolved architectural problem in generative AI. The risk escalates with agentic AI workflows as injections can trigger real-world actions. Fogel recommends machine-speed monitoring, real-time containment, and stronger identity controls until models and runtimes enforce reliable privilege separation."