Jev
Named by 10 kept items in this workspace.
5 items name it under tools, 2 under affected products, 2 under classification.key entities and 1 under classification.technologies.
Search every workspace for Jev. It matches the name as written, so it can find other things that share it.
The items that name it 10
TypeSafe AI classifier used as a judge for prompt injection and tool risk; shown to be manipulable by counterfeit policy attacks.
the page's name for the evaluated system; its prompt-injection detection rate dropped from about 83% to about 4% in Datadog Security Labs tests
TypeSafe's 'type one thinking' decision model, tested on identity resolution across scattered accounts.
TypeSafe AI model that returns structured decisions (Choice, Score, Noul primitives) with calibrated probabilities; tested by Datadog as a prompt injection classifier and tool-risk judge
Classifier architecture that takes context, a question and a list of options and outputs a probability for each option; used to build a vulnerability scanner that asks true/false questions per CWE. Input costs $0.042 per million tokens, output is free, each call takes about a quarter of a second, an
Dedicated decision model that maps unstructured language to probability distributions over finite choices, shown to be susceptible to context-based flipping of decisions
typed-question judgement API (api.typesafe.ai) that returns probabilities, choices with distributions, or scores instead of prose
All technologies in this workspace.