AIPRM for ChatGPT
ojnbohmppadfgpejeebfnmnknjdlckgj
Risk Score
5.29
Risk Level:
Medium
Recommendation:
🟡 MEDIUM RISK — review
Top Risks
- Privacy policy admits data collection + 3rd-party sharing but is NOT scoped to this extension — worst-case privacy disclosure.
- No CSP combined with 4 innerHTML DOM-XSS sinks elevates code-quality risk significantly.
- brand_mention flags 'chatgpt' as impersonation (confirmed_owner=false); name leverages OpenAI brand without verification.
- webRequest + declarativeNetRequestWithHostAccess on ChatGPT host enables full interception of AI conversation traffic.
- 12 external JS hosts contacted including test/dev endpoints; broad network surface with no CSP guard.
Evidence
- privacy_policy_scope_fail api Policy fetched but scope_extension=false, data_collection=true, third_party_sharing=true → D clause: +10.0 privacy score.
- no_csp_with_innerhtml_sinks crx csp_present=false AND 4x dom_sink_innerhtml_userctrl findings → +2.0 code quality per FIX B.
- brand_impersonation store brand_mention.is_impersonation=true for 'chatgpt'; confirmed_owner=false; not verified publisher → +2.0 reputation.
- high_permissions_no_csp manifest webRequest + declarativeNetRequestWithHostAccess on chat.openai.com; MV3 but no CSP declared.
- broad_network_surface crx 12 external JS hosts including dev/test endpoints (aiprmtest1.com, devtest-www.aiprmtest.com, github.com).
- ai_content_processing store AI extension with content_scripts on chat.openai.com; processes ChatGPT page content in-browser.
- no_developer_name store developer_name is empty string; increases accountability gap.
- featured_by_google store is_featured_by_google=true; partial trust signal but does not override privacy/code findings.
Permissions Breakdown
- activeTab low Scoped to user-invoked tab; limited blast radius.
- contextMenus low UI addition only; no data access.
- declarativeNetRequestWithHostAccess high Can intercept/redirect network requests on permitted hosts — HIGH capability.
- identity low OAuth token access; no broad scope declared.
- scripting medium Programmatic script injection into pages; elevated risk paired with host access.
- storage low Local extension storage only.
- webRequest high Observes all network requests on permitted hosts; traffic interception capability.
- https://chat.openai.com/* medium Host permission scoped to ChatGPT; content scripts run here — AI session data in scope.
Pillar Scores
Permissions6.50
Reputation5.50
Network5.50
Webstore5.50
Maintenance0.00
Privacy10.00
Code Quality4.00
CVE Exposure0.00
Scoring History
| sssiedna797ae80dp727562726963xsx | 5.27 | Medium | review | 2026-08-30 |
| v3.6 | 5.29 | Medium | review | 2026-06-16 |
Bookkeeping
Rubric v3.6
Scored at 2026-06-16 08:01
Listing SHA
2d427b669492…
Force block
— not fired
Score recovered
no
Elapsed
30.2s