NewsAnalysis
State groups and criminals build AI into the whole attack lifecycle
Recent weekly recaps describe a Russian state-linked group using Claude to modify and redeploy malware automatically when detected, North Korea's Kimsuky generating phishing decoys with local models (Ollama, GPT4All, Msty), a swarm of OpenAI agents publishing thousands of malicious packages to RubyGems in May and June, a Windows botnet with a module that drains victims' paid AI credits, and a China-based studio running 20 dating apps with 4,700 AI personas that deceived more than 25,000 people.
Why it matters
Attackers iterate at machine speed now; fixed signatures and slow human adversaries are assumptions to retire.