Securing Claude
Protect Customer Data in Claude, Claude Code and Claude Enterprise
Claude shows up in a company in three places: claude.ai in the browser and the desktop app, Claude Code in developers' terminals and IDEs, and Claude Enterprise with its connectors, Cowork and Claude in Slack. Business plans (Team, Enterprise, API) aren't used for model training under Anthropic's commercial terms; consumer accounts depend on the user's privacy settings. Check Anthropic's current terms.
The data risks differ per surface. In chat it's pasted customer records and uploaded files. In Claude Code it's secrets in .env files, database URLs and customer exports that the agent reads while it works. In Claude Enterprise it's everything people send from the web, desktop and mobile apps.
What the employee types
Draft a reply to Ada Lovelace about her
refund of CHF 2,400 to card
4532 7781 0021 9047
IBAN CH93 0076 2011 6238 5295 7
What the AI provider receives
Draft a reply to mf_mask_170b about her
refund of mf_mask_9c2e to card
mf_mask_41af
IBAN mf_mask_d7e0
Why the usual approach falls short
Most tools cover one Claude surface. Maskflare covers all of them under one policy: the browser and desktop app through the desktop agent, Claude Code through hooks, and Claude Enterprise through Anthropic's own Inference Hooks and Compliance API, with nothing to install on the device.
The Maskflare approach
On managed devices, the desktop agent masks personal data and secrets in claude.ai and the Claude desktop app, scans uploads and inspects answers, and tells your Team or Enterprise organization from personal accounts. For developers, the agent installs Maskflare hooks in Claude Code that block prompts containing secrets and mask shell, file and MCP tool input and output. For Claude Enterprise, connect Maskflare as an Inference Hooks endpoint to check every prompt in real time, including on mobile, Cowork and Claude in Slack, and connect the Compliance API to audit conversations after the fact.
Benefits
claude.ai and desktop
Prompts and uploads masked, answers inspected, personal accounts blocked or masked more strictly.
Claude Code
Hooks installed by the desktop agent block secrets and mask tool input and output.
Claude Enterprise, real time
Inference Hooks check each prompt before Claude answers, on web, desktop, mobile, Cowork and Slack.
Claude Enterprise, audit
The Compliance API connector reviews conversations and alerts on sensitive data, without storing the content.
MCP
MCP servers Claude connects to are masked and governed per group, with access requests.
Built on
Understand the underlying risks
Frequently asked questions
Does Anthropic train on Claude Team or Enterprise data?
Under Anthropic's commercial terms, business plans and the API aren't used for training by default. Consumer accounts follow the user's privacy settings. Check Anthropic's current terms.
How does Maskflare protect Claude Enterprise without an agent?
Through Anthropic's Inference Hooks: Claude sends each prompt to Maskflare before answering, and Maskflare allows, masks or blocks it. The Compliance API connector adds after-the-fact review.
Does it work with Claude Code?
Yes. Maskflare hooks block prompts that contain secrets and mask Bash, web, file and MCP tool input and output. The desktop agent installs and maintains them.
Is Claude mobile covered?
Claude Enterprise traffic from the mobile apps is covered through Inference Hooks and the Compliance API. Consumer mobile apps on personal phones are out of reach.
Can we block personal Claude accounts?
Yes. On managed devices Maskflare recognizes your Claude organization and can block or strictly mask personal accounts.
Is Claude GDPR compliant?
It depends on your plan and contract; see our Claude GDPR article. Masking keeps personal data out of prompts whichever plan you use.
See it on your own data
Try Maskflare free.
Bring your hardest prompts.
28-day free trial or 10,000 requests, whichever comes first. Point one app at the gateway or enroll a few laptops,
and watch what gets masked. Or talk to us first.
- SOC 2 Type II: In progress
- ISO 27001: In progress
- EU hosting by default · on-premises available
- Prompts and files never stored
- Never used to train AI models
- GDPR · nDSG · DPA available
- Trust center →