Publish policy bundles
Turn approved rules into compact bundles Accord Guard can enforce in the browser.
Your company already has an AI policy.Accord makes it real.
Accord turns policies into browser-side guardrails for ChatGPT and other AI tools. Employees keep moving. Admins get the signal they need.
Privacy-first by default. No raw prompt or response storage.
Built for teams handling sensitive data
Financial servicesHealthcareLegalHR teamsProduct teamsSecurity
Product
Policy enforcement for the places work actually happens.
Protect the moment
Warn, redact, rewrite, or block before risky data leaves the employee workflow.
Keep visibility private
Track policy outcomes and trends without storing raw prompts or responses by default.
Browser-first
Runs where employees already use AI
Metadata only
Policy events, not conversation archives
Policy bundles
Admin-approved rules shipped to Accord Guard
Local redaction
Sensitive values stay protected in the session
Dashboard
The signal, without the transcripts.
Usage, policy decisions, and risk levels in one place — built on metadata and redacted evidence, never raw conversations.
Northstar Financial / Overview
Supabase liveExtension live
AI requests
1,284
Active users
63
Policy events
112
Blocked
9
Guarded traffic
System status
SupabaseConnected
Extension telemetryReceiving events
Raw content storageDisabled
Risk breakdown
112events
Low61 · 55%
Medium22 · 20%
High17 · 15%
Critical12 · 10%
Provider usage
ChatGPT
Claude
Gemini
Copilot
Recent governed activity
| Time | Source | Risk | Action | Status |
|---|---|---|---|---|
| 10:46 PM | Chrome extension | medium | Blocked before AI | Blocked |
| 10:44 PM | Accord Chat | low | Allowed and logged | Resolved |
| 9:31 PM | Chrome extension | critical | Redacted and sent | Logged |
Policy authoring
Bring your policy. Ship real rules.
Upload a policy document, review extracted rules, and publish a bundle your extension can enforce. No guessing. No giant policy PDF sitting unused.
Current Guard bundle · v12
Deployed to Accord Guard
Draft
2
Pending
1
Published
7
Published rules
7Redact client identifying information before external AI use
redact_client_identifying_information
approvedv1bundle v12
Do not submit secrets and credentials to unapproved AI
do_not_submit_secrets_and_credentials
approvedv1bundle v12
Require approval for confidential data in AI workflows
require_approval_for_confidential_data
approvedv1bundle v12
1Upload your policy
External AI Usage Policy.pdf
pdf“Employees must not submit client names, account numbers, or medical records to personal AI services…”
§ 4.2 · Client information
2Accord parses it into rules
Extracted draft rules
2 of 7Redact client identifying information
draftredact_client_identifying_info
cited § 4.2Block medical records in personal AI
draftblock_medical_records_personal_ai
cited § 4.2Every rule carries its source citation and an employee-facing explanation.
3Approved rules deploy as a bundle
Bundle v12
7 rules7d197c1d54a5e4613d375f72
Live in Accord Guard
Enforced in the browser on every prompt — warn, redact, require approval, or block.
Privacy
Governance without conversation surveillance.
Accord records the facts a governance team needs: rule, action, risk level, destination, and time. Raw employee conversations stay out of the dashboard by default.
Local redaction
Protected values are replaced before submission.
Redacted previews
Teams can review outcomes without reading full conversations.
Bundle history
Every published ruleset is versioned.
Audit trail
Policy events are tracked for reporting.
Frequently asked questions
Does Accord replace the AI tools employees use?
No. Accord Guard works in the browser so employees can keep using familiar AI tools while company policy is enforced around the interaction.
Do admins read employee prompts?
No by default. Accord is built around policy metadata, actions, risk levels, and redacted previews instead of raw conversation storage.
Can we use our own AI policy?
Yes. Teams can upload policy documents, review extracted draft rules, approve what matters, and publish a bundle for the extension.
What happens when a prompt violates policy?
The rule decides the action. Accord can warn, redact, rewrite, require approval, or block the request before it reaches the AI surface.