4:32 PMTechCrunch
OpenAI unveils Lockdown Mode to protect sensitive data from prompt injection attacks
Summary
OpenAI introduced Lockdown Mode for ChatGPT to reduce the chance that sensitive data is exposed through prompt injection attacks. The feature aims to limit risky model behaviors, though it does not fully eliminate prompt injection risk.
Why it matters
Lockdown Mode lowers the operational risk of using LLMs with sensitive data, but it also formalizes that prompt injection remains an unsolved, managed threat.