Prompt injections, malicious commands that trick large language models into harmful actions, have long been a tool for attackers. Now, researchers from Tracebit have shown that defenders can use the same technique to protect sensitive data.
By placing prompt injections next to secrets such as passwords and cryptographic keys on Amazon Web Services, they can cause attacking AI agents to shut down. The injection directs the LLM to violate its own safety guardrails, forcing it to stop rather than exfiltrate data.
Comments