Article
CommunityJev Guides & Articles
Rogue Security: what Jev means for agentic security
Rogue Security's post uses Jev and System One models to frame agentic security as a decision problem: the same send_file action should be allowed, blocked or reviewed depending on who requested it and where the data goes, illustrated with policy case files rather than product results.
- Format
- Article
- Reading time
- 3 min
- Published by
- Community
- Author
- Rogue Security Research
- Last verified
- Oct 2026
Highlights
- Rogue frames the core question as: an agent can explain its next action, but that does not tell you whether it should be allowed to take it.
- Its decision trace shows one action needing a security judgment - allow, block or review - depending on authorization and destination.
- Illustrative cases: an internal finance destination allows the send; an unapproved external address blocks it; a missing recipient approval routes it to review.
- Rogue says it has a Jev-class decision model for agentic security, used for prompt injection, tool abuse, coding-agent actions and policy violations.
- It argues fast decisions make judging every step practical, not just the opening prompt, with benefit measured in coverage and interruption rates.
Watch out
Copy reviewed against the published post on 2026-10-17; the digest text was truncated at the closing section on what security teams should evaluate. The post is vendor-authored by Rogue Security, which is building a Jev-class decision model for agentic security, and its ALLOW/BLOCK/REVIEW case files are labeled as illustrative policy outcomes, not product results.
Questions about this entry
Short answers drawn from the linked project or page.
Rogue Security uses TypeSafe's Jev release to explain System One models and to argue that agentic security is a decision problem: an agent can explain its next action, but that does not tell you whether it should be allowed to take it.
Typed decisions within a defined set of possibilities, accompanied by probabilities that make uncertainty usable, rather than an open-ended string that a program has to parse.
Because the security question is bigger than whether a proposal contains a suspicious phrase. Who requested the upload, whether the destination is approved, whether the task requires the records and whether the user authorized it all matter, and a keyword cannot settle those.
Allow when the request is authorized and the destination is internal; block when a document tries to authorize disclosure to an unapproved external address; and review when recipient approval is missing. Rogue labels these as illustrative policy outcomes.
A model sounding certain is not enough; its expressed uncertainty needs a meaningful relationship with how often its judgments are correct across many predictions. That does not guarantee an individual answer, so low confidence should pause an action rather than be read as permission.
Rogue says it has a Jev-class decision model for agentic security, where Jev-class describes the role of fast, focused judgments. Its work includes recognizing prompt injection and evaluating tool abuse, coding-agent actions and policy violations.
That permissions and security policy still determine what is allowed, detailed investigation still matters for hard cases, and faster decision models make continuous step-by-step assessment more practical rather than removing the need for review.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
YouTube: Kubernetes admission control with a decision model, fail-closed.
Hacker News: SafeDep Gryph plus Jev checks for compromised coding-agent behavior.
2 pointsDetecting Compromised AI Coding Agents with Jev and GryphRead the thread on Hacker NewsX: Jev and Laya on security queries
X: Testing Jev for prompt injection
More Jev resources like this
Pillar Security analyzes Jev and System One models: a smart-home demo answered three typed questions in about 200 ms with no language model in the loop, while prompt scanning, output filtering and chain-of-thought monitoring all have nothing to inspect. TypeSafe's docs say state is not treated as hostile by default.
Jev Guides & ArticlesArticle#community#press#security
Community
HiddenLayer researchers tested Jev and an in-house System One-style model: on a simulated checkout page an attacker-controlled ad made the browser agent click away from the purchase, and TokenBreak-style edits lowered toxicity scores for prompts both models had flagged.
Jev Guides & ArticlesArticle#community#press#security
Community
HiddenLayer's insight post walks security teams through its Jev tests: on a simulated checkout page Jev gave the attacker ad's button 91% against 9% for Place order, and TokenBreak edits cut one prompt's toxicity score by more than two thirds so Jev stopped flagging it.
Jev Guides & ArticlesArticle#community#press#security
Community
Ian Maurer's ThinkThen analysis argues Jev's faster and cheaper are measurable today while its better is a bet on RLCD: unlike RLHF, which pleases humans, and RLVR, which optimizes benchmarks, Jev is trained to be reliable for programmatic use, in Diogo Almeida's own framing.
Jev Guides & ArticlesArticle#community#press#explainer
Community
A BIG CHANGE analysis of the Diogo Almeida Latent Space interview: Jev's most consequential proposition is the unit of automation, a bounded judgment inside an existing process, and the launch claim that Jev can't hallucinate needs a narrow reading because schema matching does not establish truth.
Jev Guides & ArticlesArticle#community#press#explainer
Community
TWIML episode 779 features TypeSafe co-founder and CEO Diogo Almeida on Jev, the company's model for bringing fast, reliable intelligence directly into software, and why text-generating models are poorly suited to many real-world automation decisions.
Jev Guides & ArticlesArticle#community#press#podcast
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.