{
  "entry": {
    "kind": "vuln",
    "slug": "claude-code-makes-auto-mode-the-default-permission-mode-for-paid-plans",
    "name": "claude-code-makes-auto-mode-the-default-permission-mode-for-paid-plans",
    "vendor": "Claude Code on Pro",
    "description": "From 14 August, new Claude Code sessions on Pro, Max and Team plans default to auto mode, in which a separate classifier judges each tool call and only irreversible, destructive or outward-facing actions stop for approval; Anthropic says Enterprise, API and cloud-platform users follow within the month. The stated rationale is that manual review was not working: across 1,053 paid testers auto mode caught 89 percent of harmful actions against 13.6 percent for human review, and users approved 97 percent of the prompts they were shown. This is still a fleet-wide shift in default autonomy that nobody has to opt into, and users who had previously chosen a different default get a single switch prompt.",
    "website": "https://techcrunch.com/2026/08/09/anthropic-is-turning-claude-codes-auto-mode-on-by-default/",
    "repository": null,
    "documentationUrl": null,
    "license": null,
    "version": null,
    "trustLevel": "advisory",
    "capabilities": [],
    "packaging": null,
    "configFiles": null,
    "category": "agent",
    "updatedAt": "2026-08-14",
    "risk": {
      "score": 50,
      "band": "medium",
      "signals": [],
      "dangerousCombos": []
    },
    "dangerousCombos": [],
    "vuln": {
      "id": "claude-code-makes-auto-mode-the-default-permission-mode-for-paid-plans",
      "aliases": [
        "ti-076"
      ],
      "title": "Claude Code Makes Auto Mode the Default Permission Mode for Paid Plans",
      "summary": "From 14 August, new Claude Code sessions on Pro, Max and Team plans default to auto mode, in which a separate classifier judges each tool call and only irreversible, destructive or outward-facing actions stop for approval; Anthropic says Enterprise, API and cloud-platform users follow within the month. The stated rationale is that manual review was not working: across 1,053 paid testers auto mode caught 89 percent of harmful actions against 13.6 percent for human review, and users approved 97 percent of the prompts they were shown. This is still a fleet-wide shift in default autonomy that nobody has to opt into, and users who had previously chosen a different default get a single switch prompt.",
      "severity": "medium",
      "cvss": {
        "score": null,
        "vector": null,
        "version": null
      },
      "cwe": [],
      "attackType": [
        "Other"
      ],
      "componentType": "agent",
      "component": {
        "name": "Claude Code on Pro",
        "vendor": null,
        "package": null,
        "ecosystem": null,
        "catalogSlug": null
      },
      "affected": [
        {
          "range": "Claude Code on Pro, Max and Team plans from 2026-08-14; Enterprise, API and cloud platforms to follow",
          "fixed": null
        }
      ],
      "exploited": false,
      "exploitPublic": false,
      "publishedAt": "2026-08-14",
      "modifiedAt": "2026-08-14",
      "lastVerified": "2026-11-01",
      "references": [
        {
          "type": "article",
          "url": "https://techcrunch.com/2026/08/09/anthropic-is-turning-claude-codes-auto-mode-on-by-default/"
        }
      ],
      "mitigation": "Decide this default deliberately instead of inheriting it. If your policy requires per-call approval, set permissions.defaultMode in managed settings before the change reaches your Enterprise or API tenants, and alert on any device whose defaultMode moves without a change request. Auto mode beats a human approving 97 percent of prompts, but it is a classifier, not a control you can audit per action.",
      "detectedByAnomity": true
    }
  }
}