Entity
TypeSafe Inc. – Jev AI Model
TypeSafe Inc. closed an $870M round at a $7.5B valuation shortly after launching its Jev AI safety-judge model, and shipped a minor update (1.14.0) with undisclosed experimental features (SiliconANGLE/TypeSafe blog, October 9/14). A separate academic paper published the first independent adversarial robustness evaluation of Jev, noting TypeSafe acknowledges adversarial vulnerability without quantifying it (arXiv, October 14).
This page was written and is updated by the paper from the source articles listed below. Importance is how central the topic is to current coverage; confidence is how well-sourced the page is; mentions counts the articles that fed it. How the wiki works →
## TypeSafe Inc. / Jev (AI Model)
**Type:** Company / Product
**Status:** TypeSafe closed an $870 million funding round at a $7.5 billion valuation less than a month after launching its Jev AI model (SiliconANGLE, October 9); the company has since shipped a product update, Jev 1.14.0, including experimental features (TypeSafe blog, October 14), and an independent academic paper has published the first controlled adversarial evaluation of Jev's robustness as a safety judge (arXiv:2610.17820, October 14).
**Key Facts:**
- $870 million funding round at $7.5 billion valuation, led by Andreessen Horowitz with participation from Sequoia Capital, DCVC, and unnamed angel investors (SiliconANGLE, October 9).
- Funding round closed less than a month after Jev's product debut, reflecting rapid investor interest (SiliconANGLE, October 9).
- Released Jev 1.14.0, described by TypeSafe as including "experimental features for dreamers" (TypeSafe blog, October 14). Specific capabilities of this release and the "experimental features" were not detailed in the announcement.
- Jev is characterized in independent research as "a new kind of safety judge" that generates no text or label but instead answers developer-defined questions with a probability score for each (arXiv:2610.17820, October 14).
- TypeSafe reportedly "admits that adversarial input can change these answers but gives no numbers" on the extent of that vulnerability, according to the paper's authors (arXiv:2610.17820, October 14). This is a third-party characterization of TypeSafe's public statements, not independently verified by this wiki.
- Researchers built a test harness sending identical adversarial payloads under identical policy to Jev, LLM judges, and purpose-built safety classifiers across four public-data tasks (input moderation, injection detection, routing, and at least one other task type) to compare robustness (arXiv:2610.17820, October 14). The paper's findings/results were not fully detailed in the available abstract excerpt.
**Timeline:**
- Prior to October 9, 2026: Jev model launched (SiliconANGLE, October 9).
- October 9, 2026: TypeSafe closes $870M round at $7.5B valuation (SiliconANGLE).
- October 14, 2026: TypeSafe releases Jev 1.14.0 with experimental features (TypeSafe blog).
- October 14, 2026: Academic paper published presenting first controlled adversarial comparison of Jev against LLM judges and purpose-built safety classifiers (arXiv:2610.17820).
**Connections:** AI startup funding/venture capital trends, foundation model competition, Andreessen Horowitz portfolio, Sequoia Capital AI investments, AI safety evaluation/red-teaming research, LLM-as-judge methodologies.
**Section Relevance:** AI/tech (high), business/finance (venture funding, valuations), product coverage (model releases and benchmarking), AI safety/research (adversarial robustness evaluation).
Source Articles
- Native Probability Outputs for Clinical Triage: An Evaluation of Calibration and Decision Accuracy(source)
- Teacher Supervision, Fixed Representations and Calibration: Exact Risk Calculations for Candidate-Scoring Decision Models(source)
- Narrow Decision Models: A System-1 Architecture for Calibrated, Typed Decisions(source)
- Same Attack, Different Judge: A First Controlled Adversarial Comparison of Jev, LLM Judges, and Purpose-Built Safety Classifiers(source)
- We’re releasing Jev 1.14.0 along with experimental features for dreamers(source)
- Jev creator TypeSafe closes $870M round at $7.5B valuation(source)