{
  "id": 14629474,
  "title": "Jev and other decision models can still be tricked with prompt injection",
  "url": "https://urgent.news/2026/10/15/jev-and-other-decision-models-can-still-be-tricked-with-prompt",
  "topic": "science",
  "section": "Science",
  "published": "2026-10-15T06:28:00.000Z",
  "source": {
    "name": "The Register Science",
    "slug": "the-register-science",
    "url": "https://www.theregister.com/ai-and-ml/2026/10/15/jev-and-other-decision-models-can-still-be-tricked-with-prompt-injection/5303184"
  },
  "original_language": "en",
  "account": "Recent evaluations of Jev, a decision model from TypeSafe AI, reveal that it may still be susceptible to attacks despite claims that it cannot hallucinate and does not make type errors. TypeSafe AI's Jev, which outputs probabilities and confidence scores, can still be exploited through indirect prompt injection, according to a security firm. HiddenLayer, a security business, tested both Jev and a traditional LLM, Qwen3.8-27B, to determine their resistance to manipulation. Results showed that while Jev succumbed less often to the attacks, both models were vulnerable to clickjacking via indirect prompt injection. By adding a sponsored ad to a simulated online store, researchers were able to trick both models into clicking on buttons and taking unintended actions. This suggests that even decision models, which promise cost and performance advantages, should be approached with caution. The researchers argue that a model returning a decision remains a language model at its core and its decision translates into actions in your system.",
  "summary": "Never making type errors does not mean never making errors",
  "key_points": [
    "Jev, a decision model from TypeSafe AI, may still be vulnerable to attacks.",
    "HiddenLayer tested Jev and Qwen3.8-27B for resistance to manipulation.",
    "Both models were susceptible to clickjacking via indirect prompt injection."
  ],
  "editors_take": "The vulnerability of decision models to indirect prompt injection proves that even specialized systems remain fundamentally language models that can be manipulated into performing unintended actions within a larger system.",
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "The Register",
        "title": "Jev and other decision models can still be tricked with prompt injection",
        "url": "https://urgent.news/2026/10/15/jev-and-other-decision-models-can-still-be-tricked-with-prompt-14631686",
        "published": "2026-10-15T06:28:00.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}