Skip to content
Marketing & PR · GDPR Art. 28 · CCPA

Use ChatGPT and Claude on Client Work — Without Breaching Your Client's DPA

PortEden replaces client identifiers, customer PII, and confidential brand material with placeholders before your prompt reaches OpenAI or Anthropic. Use any AI on agency work; the model never sees the underlying client.

See pricing

Free for solo agencies · No credit card · Per-client audit log built in

Maps to
GDPR Art. 28
CCPA
HIPAA
SOC 2
EU AI Act
The Risk

What Goes Wrong Without PortEden

Your Strategist Pastes a Client Brief Into ChatGPT for a First Draft

That brief carries the client's confidential roadmap, customer segments, and pricing detail — the exact fields the MSA says you'll only use to deliver the contracted service. Sending it to OpenAI without sub-processor authorisation is the Article 28 breach the GC will find on the next audit.

Your Designer Uses Claude to Refine a Campaign From a CRM Export

The export holds the client's customer list — names, emails, segment tags, sometimes purchase history. Claude's context window now holds the client's first-party data, retained per Anthropic's policy, in violation of the CCPA service-provider terms in your SOW.

Your AM Connects the Agency Outlook to an AI Assistant

Subject lines like "Acme Q2 launch — confidential pricing" and the strategy memo itself flow into the AI. One prompt-injection or subprocessor change and your client's competitor has a clean copy of the launch plan you were paid to keep secret.

How PortEden Protects Every Client Account

Client-Confidential Material, Redacted Before It Reaches the Model.

PortEden inspects every field your AI is about to see. Client identifiers, customer PII, pricing, roadmap detail, and free-text confidential material are replaced with placeholders at the boundary — never sent to OpenAI or Anthropic.

Your data
PortEdenRedact
Your AI
Claude
ChatGPT
Copilot
Gemini
Grok
Safe
Sensitive
Redacted
Compliance Reality

What Your Client's DPA Actually Requires When Your Agency Uses ChatGPT or Claude

  • GDPR Article 28 makes the agency a processor of every client's customer data. Sending that data to a third-party AI without authorising the AI as a sub-processor — and updating the Article 28 chain — is a breach of the contract every Fortune-500 client made you sign.
  • CCPA's service-provider rules and CPRA contractor terms restrict using personal information for any purpose other than the contracted service. Cross-feeding one client's data to improve a campaign for another is exactly the use the rules forbid; AI vendor logs make that risk concrete.
  • Healthcare and financial-services clients flow you their downstream HIPAA and GLBA obligations. Most consumer ChatGPT and Claude tiers do not meet either contract — the moment your team pastes regulated content, the breach belongs to you, and the indemnity letter goes back to your client.
  • Procurement reviews now ask for an AI tool inventory, a sub-processor list, and per-client data-flow maps. "We sometimes use ChatGPT" is not the answer that closes a renewal.
GDPR Article 28 — Processor
The Solution

Built For Agencies

Client-Confidential Redaction in Gmail and Outlook

Replaces 50+ identifier types — client names, customer PII, pricing, roadmap detail, and free-text confidential material — with placeholders in under 200 ms before the prompt leaves your perimeter. The AI receives only the redacted version; client material never reaches OpenAI or Anthropic.

Per-Client Policy Isolation and Sub-Processor Mapping

Each client gets its own redaction profile, its own sub-processor list, and its own data-flow map. Healthcare clients can ride a HIPAA-aware routing while a CPG client uses CCPA defaults — without one team accidentally pasting one client's data into another's prompt.

Procurement-Ready AI Tool Inventory and DPA Trail

Per-client audit log with the prompt-level evidence procurement reviews now ask for. Drops directly into the Article 28 sub-processor list, the CCPA service-provider attestation, and the SOC-2-style annual review your enterprise clients send.

Works With Claude, ChatGPT, and Copilot — No Workflow Change

Keep using Asana, Monday, ClickUp, Figma, and the AI tools your strategists, designers, and engineers already prefer. PortEden sits in front of any AI tool — no plugin to install in your PM stack, no per-team browser extension. Solo agency to global network in days, not months.

Works With What You Already Use
Gmail
Gmail
Inbox-side redaction for AI replies on client email
Outlook
Outlook
Microsoft 365 inbox-side redaction
Slack
Slack
Strip client identifiers from internal AI-summary threads
Notion
Notion
Per-client redaction across agency wikis and project docs
With and Without PortEden

The Same Workflow, Two Very Different Outcomes

Drafting a Client Strategy Memo With ChatGPT
Without
Client name, roadmap, and confidential pricing sent to OpenAI in plain text — Article 28 sub-processor breach with no authorisation in the chain.
With
Client identifiers and confidential material replaced with placeholders before the request leaves your network. The AI drafts the memo; PortEden re-hydrates the client identity locally.
Refining a Campaign From a CRM Export With Claude
Without
Client customer list — names, emails, segments — sent to Anthropic. Cross-purpose use under CCPA, retained per Anthropic's policy.
With
Segment structure and copy reach the model; customer identifiers are placeholders. The AI refines the campaign without seeing whose customers.
Inbox Search With Copilot Across Multiple Clients
Without
Every matching email — strategy memos, asset attachments, NDAs — sent to the AI in plain text. Cross-client leakage if the search spans accounts.
With
Email content reaches the model with each client's confidential material redacted under that client's profile. The AI ranks results without crossing the client boundary.
Procurement Audit From a Fortune-500 Client
Without
AM scrambles to assemble an AI tool list, a sub-processor map, and a per-prompt data-flow trail from screenshots. Renewal slips.
With
Per-client audit log of every AI use, every sub-processor, and every data flow — exportable as the procurement pack your client's GC actually wants to see.
Multi-Account Rollout to Strategists, Designers, and Engineers
Without
Each team follows AI policy by hand; one paste-and-prompt creates an indemnity event on the largest account.
With
Agency-wide defaults; per-client and per-team overrides flow from your PM tool — same policy whether the user is on Asana, Monday, or Slack.
Try It on Your Largest Account

Five-Minute Setup. Free for Solo Agencies.

Connect Gmail or Outlook via OAuth. Spin up a per-client profile. Keep using ChatGPT or Claude exactly the way your team does today — with each client's data isolated by default.

See pricing

Frequently Asked Questions

Does using ChatGPT or Claude with PortEden satisfy our Article 28 obligations to clients?
PortEden replaces client-confidential material and customer PII with placeholders before the prompt leaves your perimeter, so the underlying personal data isn't disclosed in the clear. You still need to authorise PortEden as a sub-processor in the Article 28 chain and update your DPAs — but the end-to-end data flow is the one your clients' GCs are willing to sign on.
How does PortEden handle multiple clients with different DPAs?
Each client account has its own redaction profile, its own sub-processor list, and its own audit log. A healthcare client's stricter HIPAA-aware routing runs alongside a CPG client's CCPA defaults without mixing. Per-client policy isolation is the design point — cross-client leakage is the failure mode we built against first.
Can PortEden produce the AI tool inventory procurement teams now ask for?
Yes. The audit log exports as a per-client AI tool inventory, sub-processor list, and data-flow map — the exact artifacts the SIG, CAIQ, and TRM-style questionnaires now ask for. We've shipped the templates that drop it into a typical Fortune-500 vendor packet.
Does the per-client policy work across Slack, Notion, and Drive?
Yes. PortEden treats Slack threads, Notion pages, Drive folders, and Outlook/Gmail mailboxes as redaction surfaces under the same per-client profile. A strategist, a designer, and an engineer pasting client material across three different tools all hit the same redaction layer with the same audit consequences.
What happens when a client's regulated content (PHI, NPI) flows through?
Healthcare and financial-services clients can ride stricter profiles per account: HIPAA-aware redaction for PHI, GLBA + Reg S-P-aware redaction for NPI. The audit log keeps the regulated client's evidence trail separate from your unregulated accounts so a HIPAA audit doesn't pull in non-related-party data. Compliance with HIPAA, GLBA, and Reg S-P remains the customer's responsibility — PortEden provides the technical control.
Will PortEden change my team's PM or design workflow?
No. PortEden runs in front of your AI, not your PM stack. Your team keeps using Asana, Monday, ClickUp, Figma, Slack, Notion, and the AI tools they already prefer. The redaction layer sits between any AI tool and the data the AI is about to read.
What does it cost and how long does setup take?
There's a free tier for solo agencies. Agency pricing scales by seat — full pricing is on the pricing page. Setup is under 5 minutes for a solo agency on Gmail or Outlook + ChatGPT or Claude. Multi-office agencies typically take a half-day for SSO and per-client policy onboarding.

Keep Exploring

Ready to Use AI on Client Work Without Breaking Your Client's DPA?

Five-minute setup. Free for solo agencies. Per-client audit log and procurement-ready AI sub-processor map from day one.

See pricing

Global agency network or holding company? Talk to sales →