OpenAI has announced the implementation of new security measures for ChatGPT, designed to protect organizations from emerging threats such as prompt injection attacks and automated data exfiltration.

Lockdown Mode

The new Lockdown Mode restricts the model's capabilities, limiting interactions and reducing the potential attack surface. This mode is designed for scenarios where data security is a priority.

Elevated Risk Labels

The Elevated Risk labels flag potential vulnerabilities or suspicious behavior, allowing users to proactively identify and mitigate risks. These labels provide an additional layer of protection against sophisticated attacks.

These measures represent a step forward in data protection and the prevention of misuse of language models. For those evaluating on-premise deployments, there are trade-offs to consider; AI-RADAR offers analytical frameworks on /llm-onpremise to evaluate these options.