Check Point introduces PromptGuardX to defend AI agents against prompt injection

Check Point has introduced PromptGuardX as part of its AI Defense Plane to combat the rising threat of indirect prompt injection attacks. This technology is designed to detect hidden, malicious instructions within files, such as PDFs, before they are processed by AI agents.

As AI agents increasingly handle sensitive corporate tasks—including analyzing invoices, reviewing contracts, and accessing internal databases via APIs—they become vulnerable to instructions embedded in seemingly harmless documents. Attackers can use invisible text, hidden layers, or Unicode manipulation to trick an agent into performing unauthorized actions, such as retrieving confidential data or sending information to external services.

According to Check Point’s 2026 AI Security Report, detections of long malicious prompt payloads increased approximately fivefold between March and May 2026, accounting for nearly one percent of observed prompts by May. PromptGuardX integrates with Check Point’s threat emulation to provide runtime monitoring, access controls, and enforcement at the action level to mitigate these risks.