{"data":{"id":"6d237f83-6a57-44f6-824c-d40b7b6288a4","title":"Anthropic found a hidden space where Claude puzzles over concepts","summary":"Anthropic developed a tool called the Jacobian lens (or J-lens), which reveals a hidden layer inside Claude called the J-space that contains words the AI model is thinking about but hasn't said yet. This discovery shows that what large language models (AI systems trained on massive amounts of text to predict and generate language) actually process internally can be different from what they eventually output, giving researchers a new way to understand and control how these models work.","solution":"N/A -- no mitigation discussed in source.","labels":["research","safety"],"sourceUrl":"https://www.technologyreview.com/2026/07/09/1140293/anthropic-found-a-hidden-space-where-claude-puzzles-over-concepts/","publishedAt":"2026-07-09T20:22:28.000Z","cveId":null,"cweIds":null,"cvssScore":null,"cvssSeverity":null,"severity":"info","attackType":[],"issueType":"news","affectedPackages":null,"affectedVendors":["Anthropic"],"affectedVendorsRaw":["Anthropic","Claude","Claude Opus 4.6","Neuronpedia","Goodfire"],"classifierModel":"claude-haiku-4-5-20251001","classifierPromptVersion":"v3","cvssVector":null,"attackVector":null,"attackComplexity":null,"privilegesRequired":null,"userInteraction":null,"exploitMaturity":null,"epssScore":null,"patchAvailable":null,"disclosureDate":"2026-07-09T20:22:28.000Z","capecIds":null,"crossRefCount":0,"attackSophistication":"moderate","impactType":["safety"],"aiComponentTargeted":"model","llmSpecific":true,"classifierConfidence":0.92,"researchCategory":null,"atlasIds":null}}