Scan open-source projects for possible security bugs

Big Sleep found and reported 20 previously unknown issues in widely used open-source projects.

Done with Gemini· Advanced

What they did
Google DeepMind and Project Zero used Big Sleep, with Gemini, to inspect popular open-source projects for flaws. The agent turned up 20 new vulnerabilities, most of them in FFmpeg and ImageMagick. The writeup does not spell out the workflow or how much human checking was involved.
How it went
The 20 findings were reported under standard disclosure rules, so their details are not public yet. The story does not say whether they were all valid, how serious they are, or whether fixes have landed. It does note an earlier SQLite case where the agent tracked down the root cause of a bug that Google’s threat team had seen signs of active exploitation for; that issue was CVE-2025-6965 and scored 7.2 on CVSS.
Tips
Because the findings are undisclosed, you cannot verify them yourself right now. If you try this on your own project, keep the agent in a read-only mode and have a person review anything it flags before you report or submit it.

Try it yourself with Gemini

Inspect [open-source repository URL] for likely security flaws in its [language or component]. Look for memory-safety bugs, unsafe parsing, injection paths, and secret exposure. For each suspect issue, record the file path, why it looks risky, and a rough severity guess. Do not edit code, open tickets, or send reports yet. After you finish, stop and ask me to approve any write-up, disclosure, or other irreversible action.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.