Security

Your data stays yours.

We use it only to run your AI staff. We never sell it, never share it with another business, and never train AI on it.

Keys locked away
Walled off
Logged
No card numbers
Tested on every change

Six promises.
In plain words.

Written into your contract. Where software can check a promise, our tests check it every time the code changes.

Used only for you

Your data runs your AI staff and nothing else. We never sell it, to anyone.

Never shared between businesses

Every business is walled off from every other one. No one else on Lighthand can see your customers.

Not used to train AI

Under Anthropic's commercial terms, data sent through their API isn't used to train their models by default. We don't train models on it either.

You see who looked

Only you, the people you add, and our support team when you need help. Every visit from our team shows in your access log.

No card numbers, ever

Payments happen on Stripe's own secure pages. We never see or store your customers' card numbers.

Told within 72 hours

If something goes wrong, you hear from us within 72 hours, in plain words, with what to do next.

How we make sure.

The engineering behind each promise. Your IT person can check every line in our security documents.

One business, one wall

Your account is decided by our server from your sign-in, never by anything a browser sends. Every record we look up is filtered to your business, and asking for someone else's simply returns "not found". An automatic test signs in as one business and tries every page against another.

AI sees only what the task needs

Each AI request is built from your business's data alone. Card numbers, Social Security numbers and bank account numbers are removed before anything is sent. In demo mode nothing leaves our server at all.

Encrypted on the way, keys locked at rest

Connections use HTTPS, set up by the company that hosts your Lighthand (the hosting set-ups we ship insist on it). Passwords, sign-in sessions and API keys are stored only as one-way fingerprints, so nobody can read them back; a new API key is shown to you once. Keys for payments, texting, email and calendars are encrypted with a key kept outside the database. The database itself is protected by the hosting provider's disk encryption.

Sign-in that's hard to steal

Session cookies scripts can't read, sign-ins that end after 8 hours without use (3 days at most), repeated wrong passwords lock the account for a while, and optional two-factor codes from an authenticator app.

Customers can't talk the AI out of your rules

Your limits are enforced by the software, not by asking the AI nicely. A chat customer can't see other people's bookings, unlock a discount you didn't allow, or reach anyone else's data, because none of it is within the AI's reach.

Payments and agreements you can trust

Messages from Stripe and Twilio are only accepted with their secret signature. Your signed agreement carries a fingerprint of the exact text you read, so it can't be changed afterwards.

Checked on every change

Hundreds of automatic tests, plus security scans of our code and the software it uses, run every time we change anything, so a change that breaks one of these promises is caught before release.

Don't take our word for it.

These numbers come straight from the running service. They update by themselves.

Time online, last 90 days
…

Checked every 5 minutes. Time the service was off counts against us.

Last backup
…

An encrypted copy of everything, made every day and kept for 14 days.

Last restore test
…

Every week we restore the newest backup into a spare database and check it opens cleanly.

Sign-in protection
…

Wrong passwords lock the account for a while. Two-step sign-in is there for everyone.

Want the detail? See the live status page, with every part of the service and every incident.

You're in control.

Every switch is in your own settings. You don't have to ask us.

Draft mode

New accounts start here: the AI prepares, you approve. Nothing reaches a customer until you say so.

Limits

Maximum discount, how often customers are contacted, quiet hours and words the AI must never use.

Export

Download all of your data whenever you like.

Delete

Delete it from your settings: it's gone 7 days later unless you cancel, except invoices and agreements the law makes us keep. When you leave you get 30 days to export, then it's deleted within 30 days.

Keep less

Choose how long chats and call transcripts are kept: a year unless you pick anything from 30 days to two years.

Access log

See every time our team opened your account, which part they looked at, and why.

Two-factor sign-in

Add a code from an authenticator app to your password.

Staff permissions

Staff do the daily work but can't see billing, agreements, limits or keys, or export or delete your data. Remove someone and they're signed out at once.

Pause anything

Switch off any AI role at any time, and step the AI back to draft mode in one click.

If something goes wrong, you'll hear it from us.

  1. We contain it at once. Sign people out, change keys, switch off what's affected.
  2. We tell you within 72 hours. What happened, which data, what we've done, what you should do.
  3. We help you tell your customers if you need to.
  4. We fix the cause and add an automatic test so it can't happen the same way again.

Questions owners ask.

Is AI trained on my data?

No. We use Anthropic's commercial API, and under Anthropic's commercial terms, data sent through it is not used to train their models by default. How long they keep it follows those commercial terms. We don't train any AI models of our own on your data.

Who can see my customers?

You, the team members you add, and our support team when you need help. Each visit from our team appears in your access log. No other business on Lighthand can see them, and an automatic test checks that wall every time our code changes.

What if I leave?

You can export everything at any time, and delete it yourself from your settings (it happens 7 days later unless you cancel). After you leave you have 30 days to export, then we delete your data within 30 days, except records the law makes us keep, such as invoices and signed agreements. Daily backups age out after 14 days.

Do you store card numbers?

Never. Payments happen on Stripe's own secure pages. We keep only what Stripe tells us about the payment: who paid, how much, and when.

Where is it hosted?

On a cloud hosting provider chosen by the company that runs your Lighthand, named in your Data Processing Addendum along with the other companies that handle your data for us (Anthropic for AI, Stripe for payments, Twilio for texts and calls, and an email provider). Connections use HTTPS, payment and email keys are encrypted at rest, and the database relies on the hosting provider's disk encryption.

Are you certified?

Not yet. We don't have SOC 2 or ISO 27001 certification; it's planned. We're not set up for health data under HIPAA, so we don't take on clinics yet. Instead, we publish exactly how we protect your data and the tests that prove it.