AI Security Flaws: More Severe, Lower Fix Rate

AI systems have significantly more severe security flaws than traditional software. That's the key takeaway from new penetration tests. Here's the thing: 32% of all AI and large language model security findings are rated as high risk. This is nearly two-and-a-half times higher than the 13% found in other enterprise software. What's interesting is that these high-risk AI vulnerabilities also have the lowest fix rate, with only 38% actually resolved. One in five organizations even reported an AI security incident in the past year. Experts say AI is being rolled out quickly without mature security controls, leading to more serious issues. The top concern is prompt injection, where attackers manipulate the AI. The bottom line: These widespread vulnerabilities in AI could expose sensitive data and disrupt critical systems.