{"id":79742,"title":"IvyCheck – Guard against AI risks with real-time checks","tagline":"An API for ensuring the integrity and safety of your AI application","body":"**TL;DR:** [IvyCheck](https://www.ivycheck.com/) offers an API designed for classifying user- and AI-generated texts, along with extracting sensitive data. Our specialized machine-learning models ensure the safety and quality of your AI app in real-time.\n\n![uploaded image](/media/?type=post\u0026id=79742\u0026key=user_uploads/825709/1100f870-2ce1-44f4-a45f-2c1f640a2867)\n\n### What are AI risks?\n\nAI-powered interfaces such as chatbots emerge and pose **new AI risks:**\n\n* **Prompt Injection Attacks:** Attackers try to manipulate language models to execute unauthorized actions or access sensitive data. These attacks exploit the way AI interprets inputs as natural language text. Prompt injection attacks create a risk that spans from data exfiltration to influencing decision-making processes.\n* **PII Data Leakage:** Customers may include sensitive data in requests that are then sent to LLM providers. Companies must ensure that sensitive data is detected and redacted from requests.\n* **Hallucinations:** Chatbots in RAG applications use context retrieved from document stores, posing the risk of hallucinated, inaccurate, off-topic, or made-up information to users. IvyCheck evaluates whether chatbot answers are faithful to the provided context.\n\n![uploaded image](/media/?type=post\u0026id=79742\u0026key=user_uploads/825709/84a94c2d-9885-461e-a9f0-0724ccf1748a)\n\n### How does IvyCheck help?\n\nIvyCheck provides a library of checks for managing AI risks in your application. IvyCheck is:\n\n* **Flexible**: You can add checks to your LLM app with **just 2 lines of code** and protect your users and your data from prompt injection attacks, PII data leakage, and hallucinations. Works regardless of which LLM provider you use in your app.\n* **Secure**: We use customized, fine-tuned models for the checks. We self-host all our models and your data is not sent to third-party model providers.\n* **Fast**: Since the models are small and specialized, you can expect a latency under 100-200 ms. If you require even faster endpoints, please get in touch!\n\n### The founding team\n\n![uploaded image](/media/?type=post\u0026id=79742\u0026key=user_uploads/825709/6e2a1539-300a-4ab5-ac5a-46318b4e89d2)\n\nWe’re [Tammo](https://de.linkedin.com/in/tammo-rukat-34529471) (left) and [Dustin](https://www.linkedin.com/in/dustin--lange/) (right). We met at Amazon, where we were part of the Core Machine Learning team in Berlin. We worked on systems for large-scale data quality checks, missing value imputation, and search filter optimization. We’re both obsessed with delighting our customers and solving their toughest challenges! Let’s create a safer world with AI.\n\n### How do I get started?\n\nVisit [https://ivycheck.com](https://ivycheck.com/) or [book a call](https://ivycheck.com/request-demo) directly with us. We’d love to get in touch and help you make your AI app safe today!","slug":"KkA-ivycheck-guard-against-ai-risks-with-real-time-checks","created_at":"2024-03-27T14:54:09.208Z","updated_at":"2026-07-22T09:21:40.478Z","total_vote_count":5,"url":"https://www.ycombinator.com/launches/KkA-ivycheck-guard-against-ai-risks-with-real-time-checks","share_image_url":"https://www.ycombinator.com/media/?type=post\u0026id=79742\u0026key=user_uploads/825709/1100f870-2ce1-44f4-a45f-2c1f640a2867","company":{"id":26756,"name":"Midplane","slug":"midplane","url":"https://midplane.ai/","logo":"https://bookface-images.s3.amazonaws.com/small_logos/5fda9dc0196ef666a5a460e8ffbc29dddfc27396.png","batch":"Summer 2022","industry":"B2B","tags":["Artificial Intelligence","Generative AI","B2B","Productivity","Data Engineering"],"search_path":"https://bookface.ycombinator.com/company/26756"}}