Context SecurityYour teams keep AI. The identifiers stay on the workstation.
ContextShield spots personal information as employees type into ChatGPT, Gemini and Claude, and replaces it on the workstation with realistic synthetic data before anything is sent. The reply comes back with the real values. The work never stops.
A managed extension for Chrome and Edge, deployed through Intune or GPO. PDF, Word and Excel files included.
Your AI policy says one thing. The prompts say another.
Between the policy staff signed and what a busy employee pastes into a prompt, there is a gap. That is where SINs, contact details and account numbers walk out the door.
Block it
The employee opens the platform on their phone. The information leaves anyway, this time outside your visibility.
Log it
The alert arrives after the send. The incident has already happened and now has to be assessed.
Train for it
Necessary, never sufficient. Under deadline pressure, habit beats policy.
ContextShield acts at the only moment that counts: between the keystroke and the send.
What changes for your organization
Employees keep their tools. Your privacy officer and your IT team stay in control.
Work continues
No access cut off, no warning pop-up to dismiss. The model receives a coherent request and answers normally.
Nothing passes through us
Detection and substitution happen in the browser. No copy of the typed text is sent to Context Security or to any subcontractor.
A safeguard you can document
A concrete security measure your privacy officer can describe in a privacy impact assessment, with limits written to be quoted as is.
One policy to push
Your IT team deploys the extension through Intune or GPO. No agent, no package, no accounts to create seat by seat.
What the offer includes
A per-seat subscription, a fixed monthly platform fee and no per-request fees. The quote states exactly what is covered.
The ContextShield extension
For managed Chrome and Edge, on ChatGPT, Gemini and Claude. Twenty-three identifier types recognized with no configuration, including in uploaded PDF, Word and Excel files.
Implementation and support
The managed policy template, hands-on help for your IT team, and direct support in English or French from the person who knows your deployment.
Customization modulesOptional
For identifiers specific to your sector. Quoted individually.
Accountability dashboardOptional
Hosted on a server you own.
Detection updates
Delivered through the browser's distribution channel, with no work for your team.
From first conversation to deployment
Four steps, and a written answer to every question, including when the answer is no.
Demo
Twenty minutes on your own cases, with the identifiers you care about.
Written answer
What the tool covers for you, what it does not, and your security questionnaires answered in writing.
Quote
Per-seat subscription, platform fee and any modules required. The scope covered is written in black and white.
Deployment
Your IT team pushes a forced-install policy through Intune or GPO and keeps control of the configuration.
Thirty-six seconds, uncut
Typing, detection, substitution, send, reply, restore. No editing between steps.
Video description
A continuous thirty-six second screen recording in chatgpt.com. A message containing personal information is typed. ContextShield detects each identifier on the workstation and replaces it with a synthetic value before sending. The platform receives only the synthetic values. In the displayed reply, the original values are restored locally, with no round trip to a vendor.
Already under contract
A Montreal parapublic institution has chosen ContextShield. The organization's name will be shared once it gives written permission.
- Seats
- 50
- Deployment
- Microsoft Intune
- Contract
- Signed, in progress
What ContextShield is not
A vendor that promises compliance puts you at fault. These limits are written so your privacy officer can quote them as is in a privacy impact assessment.
- Legal classificationPseudonymization is reversible. Under Quebec's Law 25, the data remains personal information. Exposure is reduced; the classification does not change.
- ComplianceNo tool makes an organization compliant. ContextShield supports your obligations and documents a reasonable safeguard.
- PerimeterManaged browser only, on ChatGPT, Gemini and Claude. A personal device outside the fleet stays out of coverage.
- ImagesUploaded images are not covered. Only PDF, Word and Excel files are.
- Sector identifiersA format specific to your sector requires a dedicated customization module.
- Compromised workstationA workstation that is already compromised stays compromised. The tool acts on what is sent to the platform, not on device security.
Frequently asked questions
The questions privacy officers and IT teams ask before a demo.
Which platforms does ContextShield cover?
ChatGPT, Gemini and Claude, in managed Chrome and Edge. PDF, Word and Excel files uploaded to those platforms are intercepted before sending too. Images are not covered, and neither is Microsoft Copilot.
Does our employees' text pass through your servers?
No. Detection and substitution happen in the browser, before anything is sent. No copy of the typed text is sent to Context Security or to any subcontractor. The central platform is hosted on your own infrastructure. The central platform is hosted on your own infrastructure.
Why synthetic data instead of tags like [NAME_1]?
Because the model still needs to understand the request. A SIN becomes another plausible SIN that passes the Luhn check. The reply stays useful, and the original values are restored on the workstation.
Does ContextShield make our organization compliant?
No. No tool makes an organization compliant. ContextShield supports your obligations and documents a reasonable safeguard. Compliance remains yours.
Is pseudonymized data still personal information?
Yes. Pseudonymization is reversible, so the data remains personal information under Quebec's Law 25. Exposure is reduced; the classification does not change.
How is the extension deployed?
Through the tooling you already run: a Microsoft Intune forced-install policy or a GPO entry in Active Directory. No package to build, no agent, no accounts to create seat by seat.
Are personal devices covered?
No. ContextShield works in the browser your organization manages. A personal device outside the fleet stays out of coverage, and we would rather tell you before your privacy impact assessment.
What about an identifier specific to our sector?
A format outside the native list is handled by a customization module, built for your case and quoted individually.
Twenty minutes, on your own cases
Write to us with your context: number of seats, the platforms your staff use, the identifiers you worry about. We reply with an honest read of what the tool covers for you.
Or write directly to communication@contextsecurity.ca