Employees
AI that's safe to use at work, without guessing what's allowed.
- Chat with Claude, GPT and Gemini
- Streaming answers and conversation history
- Search across your own conversations
- A clear explanation whenever something is blocked
Secure AI for every employee
Give every employee AI.
Give every employee AI.
Keep every company secret.
Relvant checks every message for passwords, financial data, customer details and other confidential information before it reaches an AI model. Safe questions get a great answer from Claude, GPT or Gemini. Sensitive ones never leave.
- ✓ Every message checked
- ✓ Claude, GPT and Gemini
- ✓ Full audit trail
One secure place for
- Claude
- GPT
- Gemini
- Web
- Mobile, coming soon
The problem
Your people already use AI. Your data goes with it.
Employees paste real work into AI assistants: customer lists, contract terms, financial figures, source code, passwords. Once a prompt reaches an outside AI provider, it's out of your hands.
Banning AI pushes it underground. Allowing it means accepting the leaks. Relvant is the third option: give everyone AI, and check every message on the way out.
- Ban AIPeople use it anyway, on personal accounts, with no oversight at all.
- Allow everythingConfidential information leaves the company one paste at a time.
- RelvantEveryone gets AI. Every message is checked first, and every decision is recorded.
How it works
A security check before every AI answer.
It takes a couple of seconds, and it happens on every single message.
- 1
Your employee asks
A familiar chat, like ChatGPT: streaming answers, conversation history, search, and a choice of Claude, GPT and Gemini.
- 2
RelvantGuard checks it
Precise rules catch passwords, keys, card numbers and bank accounts instantly. A guard model hosted by Relvant judges the rest against your company's rules.
- 3
Safe messages get answered
Allowed messages go to the model through zero-data-retention routing. Blocked ones are stopped, explained to the employee, and logged for your admins.
If the security check can't run, for any reason, nothing is sent. Relvant fails closed.
What RelvantGuard stops
Nine kinds of company information stay in.
Credentials and secrets
Passwords, API keys, tokens, private keys, connection strings.
Financial data
Card numbers, bank accounts, unpublished results, budgets and deal values.
Personal data
ID numbers, home addresses, health details, a named person's salary.
Customer data
Client names with private details, customer lists, contract terms.
Legal and contracts
Confidential agreements, litigation, legal advice about the company.
HR and compensation
Pay, performance reviews, hiring and disciplinary details.
Source code
Proprietary code and the internals of your systems.
Strategy and M&A
Unannounced plans, acquisitions, launches and pricing strategy.
Security and infrastructure
Internal hostnames, network layouts, vulnerabilities.
Built for the whole company
Everyone gets what they need.
Admins
Full visibility and control over how the company uses AI.
- Analytics: usage, blocked messages, spend
- Reports with CSV export, blocked text included
- An append-only activity log of every action
- Users, roles, locations and departments
Managers
See how your team uses AI, while respecting their privacy.
- My team: usage for the departments you manage
- What was blocked, by category
- Spend and activity per person
- Never the text of anyone's messages
The platform
Everything to roll out AI responsibly.
Guarded chat
A ChatGPT-style workspace where every message passes RelvantGuard first.
Analytics
Messages per day, blocked rate, active people and AI spend, by department, location and model.
Reports
The activity log, blocked messages in full, and usage per person, downloadable as CSV.
Activity log
Every sign-in, message, decision and admin change, recorded and never edited.
Locations and departments
Model your company's offices and teams, and give each department a manager.
Invitations and roles
Invite colleagues by email. Owners, admins and employees each see what they should.
Two-factor sign-in
Authenticator apps protect every account, on the web and on mobile.
Your own rules Coming soon
Add your client names, project code names and custom rules, and test them before they go live.
Usage limits Coming soon
Daily, weekly and monthly limits by spend, messages or tokens, per person or company-wide.
Security and privacy
Designed so the check happens before anything leaves.
Relvant sits between your people and outside AI providers. Here is exactly what that means.
- Checked on infrastructure Relvant controlsThe guard model runs on Relvant's own servers in the United States, so a sensitive message is read only there.
- Fails closedIf the check can't run, the message isn't sent. There is no fallback that skips it.
- Zero data retentionAllowed messages are routed only to AI endpoints that don't keep your data.
- Encrypted conversationsEvery conversation is encrypted at rest and belongs to your company.
- Everything on recordAn append-only activity log shows who did what, when, and what was stopped.
- Honest about limitsNo filter is perfect. Relvant lowers the risk with layered checks and a full audit trail. It doesn't promise that leaks are impossible.
Questions
Frequently asked questions
Which AI models can we use?
Claude Sonnet 5 and Opus 5, GPT-5.6 Sol and Gemini 3.8 Flash today, each chosen because it's available with zero data retention. Employees pick the model for each conversation.
What happens when a message is blocked?
Nothing is sent to the AI model. The employee sees which kind of information was found, so they can rephrase without it. The decision is logged, and admins can review the message in Reports.
Who can read our conversations?
The person who wrote them, your company's admins, and Relvant's support administrators. Department managers see usage numbers for their teams, never message text. Conversations are encrypted at rest.
Where is data processed?
Relvant and its guard model run in the United States. Allowed messages go to the model you choose through zero-data-retention endpoints.
Can RelvantGuard miss something?
Yes. No filter catches everything, and someone determined can reword a secret. Relvant is built for accidental and casual leaks: layered checks stop them before they go out, and the audit trail shows your admins what was attempted.
How do we get started?
Create your company's account, invite colleagues from Settings, and set up your locations and departments. Everyone can start chatting straight away.
Give your company AI it can trust.
Set up your company in a minute, then invite your team.