Hackers Are Using AI Chatbots to Break Into Companies
Your private data could be at risk as criminals bypass AI safety rules.
Researchers at CrowdStrike analyzed four previously unreported threat groups that ran AI-powered attacks between May and August, using mostly older versions of Anthropic's Claude and open-weight Deepseek models. All four circumvented AI safeguards, often by convincing models the activity was legitimate and authorized—such as penetration testing—and some used multiple models to exploit differences in guardrails.
One group, operating in Spanish and English, targeted Argentine health care entities and compromised at least 15 organizations; CrowdStrike researchers saw evidence it stole credentials and medical records from at least one victim. A Chinese-speaking group compromised government targets in Brazil and Colombia. CrowdStrike's Adam Meyers said the attacks did not use new exploits or novel vulnerabilities—they took advantage of AI's speed and volume against preventable security hygiene issues.
- Hackers are using AI chatbots to automate attacks, making them faster and cheaper.
- They trick AI safety filters by pretending the hacking is authorized, like a security test.
- One group stole medical records from 15 organizations, showing real-world harm.
Why It Matters
Your personal data is more vulnerable as AI makes it easier for criminals to attack.