These 4 critical AI vulnerabilities are being exploited faster than defenders can respond

Security researchers warn four major AI vulnerabilities have no known fixes and are being actively exploited. Prompt injection attacks succeed against 56% of LLMs, while threat actors can poison training data for just $60. Deepfake video calls have stolen tens of millions, and autonomous AI agents are being hijacked for cyberattacks. Model repositories contain hundreds of thousands of malicious files, creating an unprecedented security crisis as adoption accelerates.