Security and data handling
Our software runs in private environments you control. Every build keeps a record of what it did, and a person at your firm signs off on anything that matters.
Where your data goes
AI runs in private, client-controlled environments. Client data does not go to public AI services.
Inside your environment
Outside
Your data does not go to public AI services.
Five layers of control
Who supplies the model
We contract directly with Anthropic. The terms are commercial, backed by a data processing agreement, with no reseller in between.
Deployment settings
- Single sign-on
One login your firm already manages.
- Role design
People see only what their role needs.
- Retention settings
Set to your policy, not the defaults.
- Audit logging
A record of what the software did.
- Spend limits
A ceiling on what the AI can cost.
- Network allowlisting
Only approved addresses can connect.
Frameworks we work to
- NIST CSFWork mapped to the framework
- NIST AI RMFWork mapped to the framework
- SOC 2We led a program for a client
- AnthropicDirect commercial terms and a data processing agreement
Limits we state plainly
- Zero data retention
Retention can be set to zero through the API. The chat interface does not offer it.
- Security review
Each build gets a named review before launch.
Where does our data go?
It stays inside the private environment you control. The AI model runs there too, along with the record. Nothing is sent to a public AI service.
Is our data used to train the model?
Our contract with Anthropic is direct, on commercial terms with a data processing agreement. We make no claim here about training behaviour beyond that.
For keeping client identifiers out of AI tools, see Redaction for AI and court filings.
Send security questionnaires through the start page.
Tell us what you want built.
A working session is 45 minutes on one idea from your company: what to build, what it needs, and what it would cost.
- A fixed price before we build
- Your data stays in your environment
- A record of every action
- Reference calls under NDA
- NDA before the first session
- Since 2019