GitHub has unveiled a specialized AI model that can help developers find passwords and other sensitive data accidentally left in code. It analyzes context and can even recognize passwords that don't look like regular access keys.
GitHub has unveiled a specialized AI model that can help developers find passwords and other sensitive data accidentally left in code. It analyzes context and can even recognize passwords that don't look like regular access keys.
This is reported on GitHub.
GitHub already has tools to search for such data. However, they can’t always recognize a password that looks like a regular string of text rather than a key with a characteristic sequence of characters. And the new AI model can analyze a suspicious string and the code around it and determine whether passwords or other sensitive data are left behind.
The company also wants to prevent leaks before they happen. To do this, the company is testing code validation when it is uploaded to the repository. If a developer accidentally leaves a password, the system can detect it before it gets into the project history.
The new model will eventually be integrated into GitHub Copilot, complementing the /security-review command, which allows developers to check code changes for vulnerabilities before saving them.
Some GitHub Secret Protection and Advanced Security users have already automatically received the new model at no additional charge. However, future checks will have to be paid for with AI credits.
Previously, dev.ua wrote that after GitHub Copilot switched to pay-as-you-go, some developers faced a very rapid depletion of their monthly reserves of AI credits.
