Watch: MARS Stops a Live Prompt-Injection Attack

See MARS stop a live prompt-injection attack in real time. In under two minutes, this demo shows how MARS separates instructions from data and strips hidden instructions, poisoned content, and disguised payloads before an AI tool ever reads them — neutralizing the attack before it can hijack the agent's behavior.