A developer has created a local AI firewall called Sentinel Protocol, an open source solution that aims to address the issue of OpenAI receiving raw user input with zero filtering from most AI apps. This move comes as a response to the lack of action from others in the industry to tackle the problem.
The Sentinel Protocol is designed to scan every Large Language Model (LLM) request and response, utilizing 81 security engines to block Personally Identifiable Information (PII), catch prompt injections, detect hallucinated URLs in model output, and handle Model Corruption Poisoning (MCP) for AI agents. The solution is relatively straightforward to implement, requiring only one command to start and a single line change in the SDK. The Sentinel Protocol consists of 52,069 lines of code, has 9 dependencies, and is released under the MIT license. Notably, it operates entirely on the user’s machine, with no cloud or telemetry involved, ensuring that user data remains private and secure.
The creation of the Sentinel Protocol is likely to have a significant impact on the AI industry, particularly in terms of data privacy and security. As the use of AI apps continues to grow, the need for robust security measures to protect user data will become increasingly important. With the Sentinel Protocol now available as an open source solution, developers and users can take steps to enhance the security of their AI-powered applications, and it will be interesting to see how the industry responds to this development and how it will be implemented in the future.
Leave a Reply