WASViking AI Guardian 作者： WASViking

The fastest-moving security gap of 2026 isn't a new CVE. It's the employee who pastes a customer table, a CPF, or an AWS key into ChatGPT to "just get this done faster." Most teams find out months later, in a breach report.

WASViking AI Guardian is the browser-side component of an enterprise platform that closes that gap. It runs only on the AI tools your team actually uses (ChatGPT, Claude, Gemini, Copilot, GitHub Copilot, and about a dozen more), watches what people are about to paste, type, or upload, and asks a local WASViking Sentinel agent whether company policy allows it. If not, the action is stopped on the host before anything leaves the browser. There is no MITM, no certificate to install, no proxy to maintain.

The agent classifies content on the device in milliseconds. It recognizes personal data (including the canonical Brazilian identifiers that most US-built tools miss), health information, the LGPD and GDPR special categories, the credential families that show up in real incidents, source code, and your own organization's internal markers. What it actually persists is small: timestamps, classification labels, a short cryptographic fingerprint, and masked samples for audit. The prompt itself and the file bytes are never stored.

Policies are written in plain English in the WASViking portal, not in the browser. "Block any paste that contains a credential into an unsanctioned tool." "Warn before sending a spreadsheet of customers to any AI." Decisions are enforced on the endpoint and audited in a dashboard your DPO can sign off on.

For IT, deployment is the boring part: push the extension through Chrome, Edge, Firefox, or Brave enterprise policy alongside the Sentinel agent. Uninstall is gated by an organization-level master password, so a single curious admin cannot quietly disarm protection.

This extension is the sensor. The policy engine, the audit storage, the compliance reporting (LGPD, GDPR, HIPAA, PCI DSS, SOC2, ISO 27001, NIST AI RMF, CPRA, PIPEDA), and the operator dashboards live in your WASViking deployment. It does nothing on its own until you pair it with the agent and a subscription.

If you're piloting an AI usage policy and want evidence it is working, that's the use case.