1How client data is protected
- Isolation by default. Every record in our system is scoped to a single client at the database level. A signed-in user cannot reach another client's data, and one client's knowledge never enters another's agent context.
- Encryption. Data is encrypted in transit and at rest. Client-held API credentials are encrypted with a key stored outside the database, so a database copy alone does not expose them.
- Least privilege. Agents receive only the system access their agreed actions require, and nothing more.
- Approval gates. Money movement, irreversible record changes, and mass outbound communication wait for a human.
- Input screening. Messages reaching an agent are screened for prompt injection, credential probing, and abuse before they reach a model.
- Audit logging. Agent actions are logged with timestamps and outcomes, and logs are available to the client.
- No model training. Client data is not used to train foundation models.
- Payment data. Card details are handled by Stripe and never stored on our systems.
2Reporting a vulnerability
If you believe you have found a security issue, tell us before telling anyone else. Email support@block9.app with the subject line “Security”, or call 636-224-8069. Include what you found, how to reproduce it, and how we can reach you.
We will acknowledge within two business days, keep you updated while we work, and credit you if you would like. We will not pursue legal action against anyone who reports in good faith, avoids privacy violations and service disruption, and gives us reasonable time to fix the issue before disclosing it.
3Please do not
- Access, modify, or delete data belonging to anyone but yourself.
- Run automated scanning that degrades the service for others.
- Use social engineering against our team or our clients.
- Disclose the issue publicly before we have had a reasonable chance to address it.
4If something goes wrong
We run a defined incident process — detect, contain, communicate, fix, review — and the communication step is not last on purpose. We tell you what we know before we have the full picture, because a client who hears late has already lost the ability to act.
- Acknowledge quickly — within a few hours of becoming aware, not after we have finished investigating.
- Contain first. We will disable a capability, force escalation, or take an agent offline if that is what limits the damage.
- Give a next-update time with every message, and meet it.
- Be factual rather than reassuring. We would rather tell you something uncomfortable early than something comforting that turns out wrong.
- Write it up afterwards — what happened, what changed, and what prevents a repeat — and confirm with you that things are stable.
Severity drives pace: minor answer-quality issues are fixed in the normal cadence; repeated failures get contained and communicated the same day; an outage, a privacy issue, or a financial or irreversible action error gets immediate containment and a phone call.
5Honest limits
We are a small company running on established infrastructure. We do not currently hold SOC 2 or ISO certification of our own, and we do not carry professional liability, cyber, or E&O insurance. We will not claim otherwise, and you should not assume it. The model providers and infrastructure vendors we build on maintain their own compliance programs; ours are the controls described above. If your industry requires formal certification from every vendor, tell us early so we can be straight with you about fit.