OpenAI Introduces Lockdown Mode to Protect Against Prompt Injection — What You Need to Know

Colleagues, a cyber security update: OpenAI has introduced Lockdown Mode to offer additional protection against prompt injection.

- What it does: disables live web browsing (cache-only), blocks fetching and rendering of web images, and disables deep browsing and agent modes.
- Limitations: not foolproof — malicious instructions may persist in cache or uploaded files and still influence responses.
- Availability: currently for ChatGPT Business and select personal accounts.

Why it matters: reduces risk of sensitive data leakage when working with LLMs.

Do you think we should deploy this mode in the corporate environment?

#cybersecurity #AI #privacy #OpenAI