AI-Powered Cyberattacks
AI-powered cyberattacks are attacks in which criminals use artificial intelligence to work faster, at larger scale or more convincingly, for example by writing phishing, cloning voices, finding vulnerabilities or automating intrusions.
Also called: AI cyberattacks, AI-enabled attacks, offensive AI, AI-assisted hacking
What are AI-powered cyberattacks?
The term covers two different things:
- Attacks that use AI as a tool, such as AI-written phishing, deepfakes and automated hacking. This is the more common meaning, and the focus here.
- Attacks on AI systems themselves, such as prompt injection, data poisoning and model theft. MITRE's ATLAS knowledge base catalogs these.
How attackers use AI
Social engineering, where the impact is clearest today. The FBI has warned that criminals use generative AI to make fraud more believable and run it at larger scale:
- AI-generated phishing with perfect grammar, personalized to each target
- Deepfake video calls and cloned voices impersonating executives
- Fake identity documents and profile photos, and synthetic identities
Technical attacks, a growing area:
- Reconnaissance: quickly researching targets, employees and exposed systems
- Vulnerability discovery and exploit writing: AI can help analyze code and turn newly published flaws into working exploits faster, shrinking the time defenders have to patch
- Malware development: helping less skilled criminals write or modify malicious code, and varying it to avoid detection
- Automated intrusion: AI agents that chain steps together, such as scanning, exploiting and moving through a network, with less human effort. Major AI companies have reported disrupting state-backed and criminal groups using their models this way.
- Password attacks: smarter guessing based on patterns in leaked passwords
What's hype and what's real
AI hasn't created fundamentally new attack types. It makes existing ones cheaper, faster and more convincing, and lowers the skill needed. For small businesses, that means more attacks of better quality, especially phishing and fraud.
How to defend
The fundamentals still work, and matter more:
- Phishing-resistant [MFA](/glossary/multifactor-authentication/) everywhere
- Independent verification of payment and access requests
- Fast [patching](/glossary/patch-management/) of internet-facing systems, since the window between disclosure and exploitation is shrinking
- [EDR](/glossary/endpoint-detection-and-response/) with 24/7 monitoring, increasingly AI-assisted itself. See AI cybersecurity.
- Updated [security awareness training](/glossary/security-awareness-training/) that covers deepfakes and AI-written scams
Related terms
Sources
- Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (FBI Internet Crime Complaint Center (IC3))
- MITRE ATLAS (MITRE)
Drafted with AI assistance, then edited and fact-checked by a person. How we use AI.
