SECURITY
How we protect what you send us.
Investment firms send a vendor questionnaire before the first data room opens. We have already answered the questions most reviews ask. Ask for our security pack; there is no NDA to sign first.
How we work
- Your documents are processed in our Google business account. The analysis runs only on language models from US providers: Anthropic, OpenAI and Google.
- Only what the work needs leaves your data room, into one engagement folder in our Google business account. For full diligence, that is usually the complete data room. Your deliverables come back to you. If you prefer, we work inside your own data room, under accounts you issue.
- We delete that folder within six months after the engagement ends, or sooner if you ask. The deliverables we issued and the corrections register are kept for seven years.
- For a report, we index your documents for search in that engagement's workspace only, and the index is deleted with the documents.
- Access is limited to the people on your engagement, each named to you before work starts. Our two administrators can open the folder only to investigate an incident, and every such access is recorded in our incident log.
- We use language models only under business terms that exclude training on your data.
- Every build runs in your own environment, and no other client shares it.
- Everyone who handles client data signs in with multi-factor authentication, on encrypted devices.
- Our CISO, Stefan Brunner, a CISSP with 25 years in enterprise infrastructure and security products, owns the program and leads incident response.
In the security pack
- A security overview
- Information security policy
- AI data-handling policy
- Policy on material non-public information and personal trading
- Incident response plan
- Data retention and deletion policy
- Our subprocessors
- A pre-answered vendor questionnaire
The language models read your documents to do the work. We never train our own models on your confidential data. What an engagement teaches us goes into our systems only in anonymized form, and your engagement letter says so. Each engagement keeps one folder you can audit.
QUESTIONS WE EXPECT
Which people see our documents?
The principal who signs the work and the analysts who run it, each named to you before work starts. Our two administrators can open the folder only to investigate an incident, and every such access is recorded in our incident log.
How long do you keep our documents?
For the engagement, and no more than six months after it ends, or less if you ask. The deliverables we issued and the corrections register are kept for seven years.
Is our data used to train a model?
Not without your explicit permission. We never train our own models on your confidential data, and the language models we use run under business terms that exclude training. A company brain or a custom AI model we build for you is built on your data, in your environment, for you. We use it for anyone else only with your written consent.
Which channels do you use?
Email, or the messaging channel your firm already uses, such as Slack, so the record stays in your own archive.
Where are our documents and our build hosted?
A build runs in your own environment, under your accounts. For a report, your documents sit in one engagement folder in our Google business account. On request, we send each language model provider's own terms on where it processes data.
Which language models read our documents?
Only models from Anthropic, OpenAI and Google, under business terms that exclude training. We give you each provider's retention terms on request, and our security pack lists every service that may hold your data.
What happens when an engagement ends?
A build stays in your environment, and we delete your documents within six months, or sooner if you ask.
Who owns security at Lensential?
Our CISO, Stefan Brunner. He holds the CISSP, approves every exception to our policies and leads our response to any incident.
What happens if something goes wrong?
We follow a written incident response plan, and we tell you without undue delay, and within 72 hours of confirming an incident that affects your data.
If your questionnaire asks something the pack does not answer, send it to us. Our CISO answers in writing.