By using tools like risk management frameworks, we identify and address AI risks early, saving time and cost while ensuring our clients’ AI solutions are built on a foundation of trust and reliability.
Our research into chatbot performance using traditional ML shows how input filters increasingly need tailored training data to identify security risks.
This case study shows how compound adversarial attacks can be identified using unsupervised learning to overcome limited training data.
System prompt exfiltration is among the most alarming of LLM attacks. We propose a definition to make prompt exfiltration attacks easier to identify.
Input filters are a blue teaming operation and essential to building safe, secure LLMs.
Learn how to adopt a mindset of continuous evaluation in generative AI, exploring popular benchmarks and AI red teaming methods.
Learn how to automate the evaluation and categorization of LLM attack methods so your AI red team ensures good test coverage and finds vulnerabilities.
As GenAI implementations become more prominent, it's critical to adhere to responsible AI practices to protect your brand and foster customer trust.
Learn how these five artificial intelligence design techniques build trust in highly regulated industries like healthcare.
A CIA technique called a canary trap helps us detect AI hallucination risk in large language models (LLMs) enhanced with retrieval augmented generation (RAG).
Intent classification used in concert with a large language model (LLM) and retrieval-augmented generation (RAG) system resulted in a safer financial chatbot.
Boost AI reliability by preventing AI hallucinations with WillowTree's three-pronged approach to minimize and mitigate incorrect information produced by LLMs.
Get curated content delivered right to your inbox. No more searching. No more scrolling.