--- title: Jev and other decision models can still be tricked with... description: Security firm HiddenLayer tested TypeSafe AI's Jev decision model, which the vendor claims never hallucinates or makes type errors, and found it remains... canonical: https://daily.dev/posts/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection-xgmflzjcw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Jev and other decision models can still be tricked with prompt injection | daily.dev og:description: Security firm HiddenLayer tested TypeSafe AI's Jev decision model, which the vendor claims never hallucinates or makes type errors, and found it remains... og:url: https://daily.dev/posts/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection-xgmflzjcw og:image: https://api.daily.dev/og/posts/XGmFlZjCw.png og:image:alt: Jev and other decision models can still be tricked with prompt injection og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Jev and other decision models can still be tricked with prompt injection **[The Register](https://daily.dev/sources/theregister)** · 4 min read · 0 upvotes · 0 comments ## Summary Security firm HiddenLayer tested TypeSafe AI's Jev decision model, which the vendor claims never hallucinates or makes type errors, and found it remains vulnerable to indirect prompt injection. Researchers built a simulated online store and showed a browser agent built on Jev could be tricked into clicking a malicious 'Confirm Price' button instead of following its instructions, with chained attacks potentially giving attackers full control. A similar tokenbreak-style attack also bypassed Jev's prompt verification use case. The finding applies broadly: decision models are still LLMs under the hood using a different decoding strategy, so they inherit the same manipulation risks as text-generating models. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Can decision models like Jev be tricked with prompt injection even though they never make type errors? Yes, decision models remain vulnerable to prompt injection despite not hallucinating in the traditional sense. HiddenLayer researchers showed a browser agent built on TypeSafe AI's Jev model could be manipulated into clicking a fake 'Confirm Price' button on a malicious webpage instead of following its original checkout instructions, because the model is still an LLM using a different decoding strategy under the hood. _Teams evaluating decision-model agents for safety can track emerging AI security research like this on daily.dev._ ### What is the tokenbreak attack and how does it affect AI prompt verification? Tokenbreak is an attack technique that adds extra letters to a prompt to evade classifiers, originally documented by HiddenLayer. Researchers applied a variant of it against Jev when used for prompt verification, one of its suggested applications, and were able to sneak toxic prompts past the verification process, showing the technique works even against decision-style models. _Developers building prompt-verification pipelines can follow security research like this via daily.dev to anticipate evasion techniques._ ## Similar posts on daily.dev - [Indirect Prompt Injection remains a fundamental security challenge for AI](https://daily.dev/posts/indirect-prompt-injection-remains-a-fundamental-security-challenge-for-ai-xnrdac9em) · Brave · 1 upvotes · 0 comments - [Prompt injection breaks today’s AI agents, study warns](https://daily.dev/posts/prompt-injection-breaks-today-s-ai-agents-study-warns-2hfyd6ptc) · CSO Online · 0 upvotes · 0 comments - [Zscaler finds autonomous agents succumb to IPI traps](https://daily.dev/posts/zscaler-finds-autonomous-agents-succumb-to-ipi-traps-wmyyoiejc) · InfoWorld · 1 upvotes · 0 comments - [Prompt Injection Attacks In Agentic AI Security Risks](https://daily.dev/posts/prompt-injection-attacks-in-agentic-ai-security-risks-0poqxk9z7) · Cyble · 1 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#ai-security](https://daily.dev/tags/ai-security), [#prompt-injection](https://daily.dev/tags/prompt-injection), [#jev](https://daily.dev/tags/jev) [View this post on daily.dev](https://daily.dev/posts/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection-xgmflzjcw) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Jev and other decision models can still be tricked with prompt injection","url":"https://daily.dev/posts/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection-xgmflzjcw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection-xgmflzjcw"},"datePublished":"2026-10-15T06:28:29.646Z","dateModified":"2026-10-15T06:28:52.462Z","description":"Security firm HiddenLayer tested TypeSafe AI's Jev decision model, which the vendor claims never hallucinates or makes type errors, and found it remains...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/fce799798df82dad290ac736a31fb44b?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/fce799798df82dad290ac736a31fb44b?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The Register","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The Register","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/66aa2113fdad463992ffcbf0e8963fda","url":"https://daily.dev/sources/theregister"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection-xgmflzjcw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents,ai-security,prompt-injection,jev","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The Register","item":"https://daily.dev/sources/theregister"},{"@type":"ListItem","position":3,"name":"Jev and other decision models can still be tricked with prompt injection"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection-xgmflzjcw#faq","mainEntity":[{"@type":"Question","name":"Can decision models like Jev be tricked with prompt injection even though they never make type errors?","acceptedAnswer":{"@type":"Answer","text":"Yes, decision models remain vulnerable to prompt injection despite not hallucinating in the traditional sense. HiddenLayer researchers showed a browser agent built on TypeSafe AI's Jev model could be manipulated into clicking a fake 'Confirm Price' button on a malicious webpage instead of following its original checkout instructions, because the model is still an LLM using a different decoding strategy under the hood. Teams evaluating decision-model agents for safety can track emerging AI security research like this on daily.dev."}},{"@type":"Question","name":"What is the tokenbreak attack and how does it affect AI prompt verification?","acceptedAnswer":{"@type":"Answer","text":"Tokenbreak is an attack technique that adds extra letters to a prompt to evade classifiers, originally documented by HiddenLayer. Researchers applied a variant of it against Jev when used for prompt verification, one of its suggested applications, and were able to sneak toxic prompts past the verification process, showing the technique works even against decision-style models. Developers building prompt-verification pipelines can follow security research like this via daily.dev to anticipate evasion techniques."}}]} ```