Top AI Security Incidents & Vulnerabilities
AI systems face a range of security threats, from data privacy breaches to sophisticated model manipulation, with publicly reported incidents increasing by 56.4% from 2023 to 2024.
The top 3
- Most Notorious AI Data Privacy Breaches: Notable AI data privacy incidents include the ChatGPT data leak in March 2023, the Clearview AI data breach in February 2020, and Samsung's confidential data leak via ChatGPT.
- Top AI Model Manipulation & Jailbreak Techniques: Common AI jailbreak techniques include roleplay scenarios like 'DAN' (Do Anything Now), multi-turn prompt chaining such as the Crescendo technique, and adversarial suffix attacks.
- Major Security Flaws in AI-Generated Code: Research shows 45% of AI-generated code contains security flaws, with models struggling particularly with Cross-Site Scripting (failing 86% of the time) and generating insecure cryptographic implementations in 14% of cases.
Sources
Open the full topic