{"data":{"id":"3ab09267-af7b-401d-82ff-9a7ef1a3b33e","title":"Same Attack, Different Judge: A First Controlled Adversarial Comparison of Jev, LLM Judges, and Purpose-Built Safety Classifiers","summary":"This paper presents the first controlled adversarial evaluation of TypeSafe AI's Jev, a safety judge that returns probabilities for developer-defined questions rather than text or labels. Across four tasks and fifty attack variants from thirteen families, Jev was the most robust system on input moderation, bypassed by 9% of blocks under a benign-sounding question versus 56% for GPT-6 Astra. On phishing detection, forgeries bypassed 19-29% of flagged emails, and Jev detected only 62% of phishing emails without attack, against 77-81% for the GLM judges.","solution":"N/A -- no mitigation discussed in source.","labels":["research","security"],"sourceUrl":"https://arxiv.org/abs/2610.17820v1","publishedAt":"2026-10-13T06:43:29.000Z","cveId":null,"cweIds":null,"cvssScore":null,"cvssSeverity":null,"severity":"info","attackType":["jailbreak","model_evasion"],"issueType":"research","affectedPackages":null,"affectedPackageNames":null,"affectedPackageRefs":null,"affectedVendors":["Meta","OpenAI"],"affectedVendorsRaw":["Jev","TypeSafe AI","GPT-6 Astra","GLM judges","Llama Guard 3","Prompt Guard 2"],"classifierModel":"claude-haiku-5-5","classifierPromptVersion":"v4","summaryPromptVersion":"v2","headline":null,"headlinePromptVersion":null,"cvssVector":null,"attackVector":null,"attackComplexity":null,"privilegesRequired":null,"userInteraction":null,"exploitMaturity":null,"epssScore":null,"epssCheckedAt":null,"kevDateAdded":null,"advisoryAliases":null,"affectedPackagesSource":null,"affectedPackagesCheckedAt":null,"patchAvailable":null,"disclosureDate":"2026-10-13T06:43:29.000Z","capecIds":null,"crossRefCount":0,"attackSophistication":"moderate","impactType":["safety","integrity"],"aiComponentTargeted":"model","llmSpecific":true,"classifierConfidence":0.9,"researchCategory":"preprint","atlasIds":["AML.T0054"]}}