Skip to content

Security scanning

Prompt-injection scanning with Llama Prompt Guard 2 86M, run inside Sluis. €0.01 per completed scan, no token surcharge, no scan deadline.

Security: configure prompt-injection scanning.
Security: configure prompt-injection scanning.English interface · illustrative demo data. Open the image for full size.

Prompt-injection and jailbreak detection is opt-in and runs at the gate, before dispatch. It has three modes: off | log | block.

  • off skips the scan.
  • log records findings without blocking traffic.
  • block refuses detected attacks. It fails closed when Prompt Guard is unavailable or classification fails or is incomplete. There is no automatic model fallback.

How scanning works

  • Security always uses Llama Prompt Guard 2 86M (sluis/prompt-guard-2-86m), inside Sluis, with no external transfer of scan text.
  • Overlapping windows of at most 512 model tokens preserve segment boundaries, within a total scan budget of 4096 to 65536 tokens (8192 by default).
  • The detection threshold is fixed at 0.8.
  • Long inputs use selected excerpts. A safe result covers only the inspected text, not the full input. The audit record keeps the inspected and total bytes.
  • Scanning follows data protection, which can retain personal data when it is off or logging only.
  • Scans add variable latency. They wait for the scanner until it answers or the caller disconnects, and there is no scan deadline.

Cost and audit

  • €0.01 per completed logical scan, once regardless of windows or verdict, including sampled-safe results. No token surcharge.
  • Off, preflight refusals and technical failures are not charged.
  • Each scan has a separate, linked billable audit entry: security.scan. Charges can still apply when the parent request is blocked or served from cache.
  • Workload overrides can change the scan mode.
  • Separately, optional Nemotron privacy inspection uses Nebul under your residency policy and is token-priced.

Built with Llama. Meta Llama 4 Community License.

Anomaly detection

Opt-in key-behaviour anomaly detection runs as a background job with zero request latency: per-key baselines from robust statistics with hour-of-week seasonality, plus a multivariate isolation-forest layer.

Alerts are explainable, never a bare score. They appear in the Console's Security view, optionally by email.

Account security

Choose how people sign in, add a second factor and decide what your organisation requires. You manage your own settings in Profile. Organisation-wide rules live in Settings › Application security.

Sign-in options

On the sign-in screen, people can use:

  • email and password
  • Google
  • Apple
  • a Microsoft work or school account (Microsoft Entra ID)
  • your own OpenID Connect provider, once your owner has set it up

Which buttons appear depends on your deployment and your organisation's settings. A session lasts 8 hours, and Sluis re-checks membership, single sign-on and second-factor status on every request.

Lost a device, or think someone has your session? Open Profile and choose Sign out everywhere.

Second factor for your account

A second factor protects your account if your password leaks. Add one in Profile:

  1. Open Profile.
  2. Add a passkey, set up an authenticator app, or turn on email codes. Any TOTP app works, such as Google Authenticator, Microsoft Authenticator, 1Password or Bitwarden.
  3. Save the ten one-time recovery codes that appear when you add your first second factor.
  • Each recovery code works once. Regenerating the set invalidates the old one, and you need a verified session to do it.
  • To change factors you must first prove one you already have. The last passkey or authenticator app that your organisation requires cannot be removed.
  • Five wrong email codes end that challenge. Five wrong authenticator codes lock the authenticator for five minutes.
  • Sluis emails you when a passkey is added to your account.
  • Changes to these settings are recorded in the organisation's audit trail.
Profile: authenticator-app settings for two-factor sign-in.
Profile: authenticator-app settings for two-factor sign-in.English interface · illustrative demo data. Open the image for full size.

Require two-factor authentication

Owners and admins can require a second factor from every member.

  1. Open Settings › Application security.
  2. Switch on Require two-factor authentication.
  3. Under Required method, choose passkeys, email codes or an authenticator app.
  4. Confirm the selected method for your own account to save the policy.
  • Owners and admins always need a second factor of their own, whether or not the organisation requires it. At their next sign-in, an owner or admin without one is sent to the enrol screen with a notice, and until they finish only the setup pages are available. Members only need one when the organisation requires it. The Google sign-in rule below applies to owners and admins too.
  • Members without the required proof must set it up before they can use the organisation. Until then only the setup pages are available. The settings page shows how many members still need to do so.
  • Email codes need no setup: members can verify by email at sign-in. Passkeys and authenticator apps must be set up first.
  • A session started with Google sign-in counts as verified. Apple, Microsoft and password sign-ins do not, and neither does a password sign-in to an account linked to Google.
  • A parent organisation's requirement also applies to its owners and admins when they act inside a client organisation.
  • Turning the requirement off leaves personal security settings unchanged.

Single sign-on with Microsoft

The owner can require Microsoft sign-in for the whole organisation. Admins can see the rule, but only the owner can change it.

  1. Sign in once with Microsoft, using a work account from the directory you want to allow. This prevents a lockout: you must have signed in from a listed directory yourself before you can require it.
  2. Find the directory (tenant) ID in the Microsoft Entra admin center under Overview, Tenant ID.
  3. Open Settings › Application security. On the Microsoft and Google directories card, list the directory IDs you accept, one per line.
  4. Switch on Require Microsoft sign-in for this organisation.
  • Members must choose Continue with Microsoft and use a work account from a listed directory.
  • Password, Google and Apple sign-ins are refused for this organisation. Members who signed in another way lose access on their next request and must sign in again.
  • You can list up to ten directories. Personal Microsoft accounts do not qualify.
  • People who set up a personal second factor (an authenticator app or email codes, or a passkey on a password account) cannot use Microsoft or OpenID Connect sign-in. They remove it in Profile first.
  • Temporary passwords cannot be set for members. Someone who already has an account accepts the invitation link first and then signs in with Microsoft. A new person signs in with Microsoft at the invited address instead of choosing a password.
  • The same card lists Google Workspace domains. They only decide which accounts can connect mail and calendar, see Mail and calendar.
  • Every change is recorded in the organisation's audit trail. If an organisation locks itself out, SevenLab can switch the rule off.

OpenID Connect providers

Next to Microsoft, or instead of it, you can connect your own identity provider: Keycloak, Okta, Microsoft Entra, ADFS 2016 or later, or authentik. Only the owner can change the connection.

  1. Open Settings › Application security and find the OpenID Connect sign-in card.
  2. Copy the Redirect URI shown there and register it as the redirect (callback) URI at your provider.
  3. Enter the Issuer, Client ID and Client secret. Set Button name to what people should see after Continue with.
  4. On the cloud, prove each email domain. Under Email domains, create the TXT record shown at your DNS provider, then choose Verify.
  5. Choose Test connection, then switch on Allow sign-in through this connection.
  6. To require it, sign in through the connection yourself once, then switch on Require single sign-on through this connection.
  • On the cloud, members choose Continue with SSO and Sluis finds the organisation from their work email domain. On a self-hosted Edge installation the button is Continue with followed by your button name.
  • Create accounts on first sign-in adds people when they first sign in. Turn it off to allow only people who are already members.
  • Roles follow your Group mapping. When several rows match, the highest role wins, and everyone else gets the default role. Mapping a group to owner is a separate opt-in. Update roles on every sign-in applies the mapping each time.
  • People deactivated through SCIM are not added back by signing in. Reactivate them through SCIM.
  • When single sign-on is required, members can use either a listed Microsoft directory or this connection. You cannot turn the connection off while it is required.

SAML through a broker

Sluis does not speak SAML directly. If your identity provider only offers SAML, put a broker that speaks OpenID Connect in front of it, such as Keycloak, and connect the broker as an OpenID Connect provider (see above).

  • ADFS 2016 or later speaks OpenID Connect and can be connected directly.
  • Older ADFS versions and other SAML-only providers need a broker.

Break-glass owner access

Single sign-on can lock everyone out when your identity provider breaks. Mark up to two owners as break-glass seats to keep a way in. Do this before you require single sign-on.

  1. Open Settings › Application security and find the Break-glass owners card.
  2. Switch on the owners who keep password sign-in. Each needs a password already.
  3. Make sure each of them holds a second factor: a passkey or an authenticator app.
  • A break-glass owner can sign in with a password while single sign-on is required, but only with a second factor.
  • It works for Microsoft directories and for OpenID Connect alike.
  • If an organisation locks itself out anyway, SevenLab can switch the rule off.

What keeps working

Requiring two-factor authentication or single sign-on changes how people sign in. It does not touch:

  • API keys, admin tokens and SCIM tokens. They are not sign-ins and keep working.
  • Sign-in through your identity provider, for members from a listed directory or the connected provider.
  • Break-glass owners, who keep password sign-in with a second factor.
  • Other organisations. A member of several organisations can still choose another one. The requirement applies only to the organisation that set it.

Existing sessions are checked on their next request. Those that no longer meet the requirement lose access to that organisation then. Sign out everywhere stays a separate action.

Verify the seal

Every call appends an entry to the hash chain. Each entry's hash is sha256(prev_hash + record), so any altered field breaks every link after it.

Read the audit metadata over the admin API. A full export and a chain verification run through the gateway CLI.

In the Console's Audit log, each request is one row. Expand it to see the checks it ran:

  • name recognition
  • LLM privacy inspection
  • the workload scope guard
  • the security scan

Each check is its own sealed entry in the chain, with its own cost. GET /admin/audit returns them in the request's checks list.

# audit metadata over the admin API, admin token scope audit:read
curl "https://api.sluis.ai/admin/audit?from=2026-09-01" \
  -H "Authorization: Bearer $SLUIS_ADMIN_TOKEN"

# a full export and a chain verification are operator CLI subcommands
sluis-gateway audit export --tenant $TENANT_ID
sluis-gateway audit verify-chain --tenant $TENANT_ID

# → chain intact