Enterprise
Trust center
Security architecture, privacy and data handling, availability and incident response — with compliance status stated as status rather than as a claim.
Status, not claims
Certifications held
0
Target dates given
0
There is no date on this page for an audit that has not been scheduled. The data behind it has no date field, so one cannot be added carelessly.
Compliance status
Not certified against anything. Here is what that does and does not mean.
A trust page that leads with badges is telling you what it wants you to conclude. This one leads with the absence of them, because that is the fact a security reviewer will establish within five minutes anyway and would rather not have had to dig for.
Framework status
5 frameworks · 0 certifications held| Framework | Status | The position | Mechanism |
|---|---|---|---|
| SOC 2 | not certifiedmechanism documented | No report and no audit under way. The controls a SOC 2 examination looks at — access control, change management, monitoring — are implemented and documented, and you can read how rather than read an opinion about them. | Read it |
| ISO/IEC 27001 | not certified | Not certified and not in an audit cycle. Naming a target date for one would be inventing a commitment. | — |
| ISO/IEC 42001 | not certifiedmechanism documented | Not certified. The AI management-system practices the standard asks about — an inventory with named owners, versioned policy, review — are what the product does, so they are inspectable in the product itself. | Read it |
| GDPR and DPDP | not certifiedmechanism documented | Data-protection law is not something a vendor is certified against. What we can state is where data sits, what leaves, and what is never copied — all of which is documented rather than asserted. | Read it |
| HIPAA | not certified | No BAA offered today. Saying otherwise to keep a deal moving would be the exact failure this page exists to avoid. | — |
“Not certified” is not very useful. “Not certified, and here is the mechanism” is.
3 of 5 frameworks have the thing they ask about documented and inspectable — not summarised, not attested to, written down where you can disagree with it. That is a weaker position than a certificate and a more useful one than a badge with a footnote.
Claims that can be false
Properties, not promises.
A promise is unfalsifiable — nobody can check whether we take security seriously. A property is false the moment a counter-example exists, which is the only kind of claim worth putting on a page like this. Each of these is one.
- No endpoint returns a stored credential
- At any scope, including admin, including support. There is no code path that reads one back out, which is what makes "the agent never sees the key" a property rather than a policy.
- No endpoint amends an evidence record
- Sealed before the response returns, and chained so a removed record fails verification rather than merely being absent.
- No model call in the decision path
- A policy is evaluated deterministically. The same inputs and the same rule version reach the same conclusion, which is what makes a decision replayable.
- No hold can be extended
- There is no endpoint for it. A hold that can be extended indefinitely is a queue, and a queue defaults to eventually yes.
- Tenant is part of the primary key
- Not a filter applied by application code. A filter can be forgotten in one query; a key cannot.
- Nothing read from a governed source is stored
- OpsAI records that a read happened and what it was permitted to do — not what came back. There is no copy of your data.
Falsifiable properties
6
each with a mechanism
Credential-reading endpoints
0
at any scope
Evidence-editing endpoints
0
sealed means sealed
Copies of your data
0
reads are recorded, not stored
What OpsAI will not do
Listed so nobody has to ask on a call.
Four of these come up in every enterprise conversation, usually late and usually awkwardly. Two of them are commitments about your data and two are commitments about how we describe ourselves.
Train any model on your data, your policies or your decision records.
Use your estate as a reference or case study without a written agreement.
Offer a certification we do not hold, or a date for one we have not scheduled.
Claim to prevent prompt injection. Nothing does reliably; we make it ineffective at the boundary.
Availability and incident response
No uptime figure, because we will not publish one we cannot evidence.
A percentage on a trust page is a claim about a measurement period, a definition of downtime and a monitoring system. Publishing one without naming those three is theatre, and inventing the number outright is worse.
What can be described is the shape of the response. The product itself treats detection and containment as separate events, and the same discipline applies to how we run it: the gap between the two is the figure worth reporting, because an incident notified quickly and contained slowly is worse than the reverse.
How the product records it, as an example
9 seconds
Between detected and contained in the sample estate’s illustrative incident. Two events rather than one, which is what makes the gap a number instead of a narrative.
Reporting a vulnerability
Report to security@opsai.dev. Include what you did, what happened, and what you expected. We acknowledge receipt and tell you what we intend to do about it.
Please do not test against another organization’s tenant and please do not run automated scanning against production. A sandbox tenant is available for security testing on request.
- Notification
- An incident affecting your tenant is reported to your named contact with what is known at the time, rather than held until the picture is complete.
- Your own record is unaffected
- Evidence is sealed and chained per tenant. An incident on our side cannot amend your records, because no code path can.
- What we will not do
- Publish a status-page uptime figure without stating the measurement period and the definition of downtime behind it. A number with neither is decoration.
How it is built
Security at OpsAI
Where the control points sit, how identity is attested and how credentials are held rather than handed out.
Your controls
Compliance at OpsAI
Which of your control obligations the record can answer, and which it cannot.
Where it runs
Deployment
Three options, and what leaves your network under each — including the one where nothing does.
Where to go
If this page failed your review, it failed it on the first section.
Which is the intended behaviour. A trust page that defers the difficult answer wastes six weeks of two organizations' time, and the difficult answer here is the first thing on it.