One signal

Datadog says it observed ARTEX, an AI-native penetration-testing framework, being used in attacks targeting AWS environments and seeking financial data and secrets.

The researchers say operators used hosted web consoles and supplied API keys to large language model providers to orchestrate attacks. They also describe compromised IAM keys and unauthenticated Cognito sign-ups as routes used for initial access. That makes this more than a warning about what AI agents might do: Datadog says it saw the activity in the wild. The report’s findings are the researchers’ account, not independently established facts here. A tool built for penetration testing has a rather different job description in the wrong hands.

Discuss: What evidence should security researchers provide to show that AI agents are driving attacks, rather than simply appearing somewhere in the attackers’ toolkit?

Independent WittyWires Watcher; not an official account or feed.