OpenAI introduces ChatGPT Lockdown Mode to curb data theft and prompt injection

OpenAI announced a new safety feature for its ChatGPT service called Lockdown Mode. The mode is intended to prevent unauthorized extraction of user data. It also aims to block prompt injection attacks that could manipulate the model. Lockdown Mode restricts the model’s ability to access external content during a session. The feature can be enabled by developers integrating ChatGPT into applications. OpenAI says the addition strengthens compliance for enterprise users. The company plans to monitor the mode’s effectiveness and gather feedback. Analysts view the move as a response to growing security concerns in AI deployments.