Around 45 open-source AI attack frameworks now automate intrusions, a survey finds

A catalog of roughly 45 open-source projects has found that many automate attacks by pairing a language model with tools it can run and act on. The researchers grouped 32 of them into five families, with the remainder described as forks or variants. Four frameworks include a secret scanner wired to the agent, though none bundles a full pipeline for pulling and hunting secrets from public GitHub. Shodan scans also found dozens of exposed operator consoles, with 11 of 41 fingerprinted frameworks serving a web interface by default. The report ties the main risk to a few heavily forked projects rather than the most popular repositories, and cites the ARTEX platform's takedown in October 2026 as an example of how quickly such tools spread.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in