Security · CloudSEK ·
Researchers document the first fully autonomous AI-agent cyberattack
An open-source agent framework planned, ran and reported on an attack that stole documents from Tajik government systems.
What happened
CloudSEK researchers found what appears to be the first documented case of an AI agent running a cyberattack end to end. A threat group used the open-source Hermes agent framework to plan, execute and report on attacks against Tajik government systems, exfiltrating documents.
An exposed server revealed the operation in detail, including logs of the agent’s decision-making. Researchers say the evidence does not establish whether a government directed the group.
What it means for you
Autonomous attack tooling is now off-the-shelf. Rate limits, anomaly detection and least-privilege access matter more than ever.
Sources
Written by TheEvents from public reporting and the companies’ own announcements. Facts can change as stories develop — check the sources for the latest.
Related
Binance’s $6.3 billion hack started with a fake job interview
The FBI attributes the largest crypto theft on record to North Korea’s TraderTraitor group; the entry point was a malware-laced coding assignment.
Indian CA eMudhra exits the SSL business after a private key turned out to be “4”
Browsers blocked all eMudhra certificates after researchers recovered weak intermediate private keys almost instantly.
PHP2Shell: two critical PHP flaws were exploited before the patch shipped
Critical vulnerabilities potentially affecting millions of PHP sites, found with the help of frontier AI models and attacked before disclosure.