{"data":{"id":"9631e6b8-5f87-4ea2-9e31-f1aeaeafb4ac","title":"Is Cursor safe? 6 risks, controls and best practices","summary":"Cursor's Privacy Mode is documented as keeping customer data out of model training, with zero data retention agreements across its model providers, and it is on by default for team members. The source argues that the agent's guardrails are weaker: Cursor's own documentation calls its run modes, Auto-review, the command allowlist and .cursorignore best-effort rather than a hard security boundary, with Workspace Trust off by default.","solution":"The source recommends changing the defaults this week and locking them with team policy, then adding controls the agent cannot reach: no over-permissioned credentials on the machine, egress denied by default, agent config reviewed as code, disposable environments for unknown code, and a record of every action every agent takes. It also notes that Cursor's OS-level sandbox, dotfile protection, Hooks and MDM policies exist as controls, with Hard enforcement mostly on Enterprise plans.","labels":["security","privacy"],"sourceUrl":"https://adversa.ai/blog/is-cursor-safe/","publishedAt":"2026-11-15T08:35:00.000Z","cveId":null,"cweIds":null,"cvssScore":null,"cvssSeverity":null,"severity":"info","attackType":[],"issueType":"news","affectedPackages":null,"affectedPackageNames":null,"affectedPackageRefs":null,"affectedVendors":[],"affectedVendorsRaw":["Cursor"],"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":"2026-11-15T08:35:00.000Z","fixReleasedAt":null,"daysToFix":null,"fixCheckedAt":null,"capecIds":null,"crossRefCount":0,"attackSophistication":"moderate","impactType":["confidentiality","integrity"],"aiComponentTargeted":"agent","llmSpecific":false,"classifierConfidence":0.85,"researchCategory":null,"atlasIds":null}}