Control Plane · works on its own
Permitted AI, its cost, and the evidence.
The Control Plane is Saible's service, run in Stockholm for EU organisations and in Ohio for US organisations. On its own it reads your providers' bills, your contracts and what you declare, and it has read-only access to your repositories and AI accounts. If you add the gateway later, each request your applications send through it is added to the record.
Overview
Capabilities
01AI in use and its cost
- Spend overviewCost by provider, model and day, from the providers' own billing feeds.Spend → Spend overview
- ContractsEach commitment as a burn-down: the amount consumed so far, the projection to the end of the term, and the amount that would be forfeited or exceeded.Spend → Contracts
- BudgetsEach team's spend for the month against the bill, projected at the current run rate, with thresholds at 80% and 100%.Spend → Budgets
- Landscape reportOne report on the AI in use: providers, models, spend by team and the readiness gaps, ranked.Evidence → Landscape report
02Context of each application
- ContextThe context of each application: models, prompts, tools, MCP servers, knowledge sources, memory stores and guardrails. It is read from the application's own repository and accounts and confirmed as a signed context manifest.Control → Context
- Connected systemsContext an application keeps in Confluence, SharePoint or another system, listed through that system's MCP server and only within the scope an admin sets.Control → Context
- Change eventsConnect a GitHub push webhook or an EventBridge rule for Bedrock. Each change notifies Saible, which reads the application's sources again. The nightly read remains the fallback.Control → AI catalogue → an application
- OntologyEvery business element your AI uses, where it is read from and which applications use it, with related context ranked by stated reasons.Control → Ontology
03Approval of what may run
- Application registryEvery application with its owner and its status: provisional, approved or stopped. A missing answer never widens access.Control → AI catalogue
- PolicyThe rules in force, written as sentences, in one signed policy revision proposed by one person and approved by a second.Control → Policy
- Review queueEvery change waits here for an organisation admin to decide. The person who proposed a policy change never approves it. With the gateway, the impact of a change is replayed over your own requests.Control → Review queue
- Assessment runPast usage from the bills, assessed against a versioned policy model endpoint by model endpoint, before any gateway enforces the policy.Evidence → Assessment run
04Checks before and after it runs
- EvaluationCandidate models run over an application's golden set, with each request decided by your policy and recorded. The figures are shown side by side without a verdict and sealed as a validation pack.Control → Evaluation
- ExposureWhat each AI platform's configuration permits, read with the credentials you already gave Saible. Each finding comes with the script that fixes it. Built, but not yet run against a live account.Control → Exposure
05Evidence
- RegistersWhat you declared compared with what the records show, and the parts of the DORA register, the GDPR Art. 30 record and the AI Act inventory that the records can evidence.Evidence → Registers
- ReadinessFor each article of the GDPR, the AI Act and DORA: held, partly held or not yet held, and who can close the gap. It is a map of readiness, never a score.Evidence → Readiness
- Policy packsObligations quoted from the acts with their EUR-Lex links, and the policy rules each pack adds.Evidence → Policy packs
- Evidence packThe zip file you give an auditor: assessment records, the policy behind them, registers, readiness and the record of who approved each change.Evidence → Evidence pack
06The new console, organised by control coverage
- AI catalogueEvery AI system found in the repositories of your GitHub organisation or GitLab group, on your connected platforms or at registration. Each stays unclaimed until an owner claims it and adds only its purpose and risk class. Agents, MCP servers and tools have their own registries.Catalogue → Systems
- Data mapThe personal and special category data each AI system reaches, through which application, and where the data is held. Data that is not classified is never treated as safe.A system → Map
- CoverageEach AI system's controls in seven areas: covered, partly covered or not yet covered, with the next control to add. It is never a verdict of compliance.Home · a system → Coverage
- Write-backA fix on Amazon Bedrock or OpenAI is proposed, tested in a dry run against what the platform holds, approved, applied under a separate write role and read again. By default a person applies it.Policy → Write-back
- Requests from your logsOn request, through the API, Saible reads requests from your own AI platforms' logs as metadata, never a prompt or a person, and checks them against the policy in force.A system → Requests
- Policy registryThe rules in force, the GDPR, AI Act and DORA packs, the policy each AI system receives, and rules suggested from a scan of your code.Policy → Policies
- Model labEach model's figures on your data classes and golden sets, beside its cost. The page shows figures, never a verdict. A model switch goes to the review queue.Models
- OntologyYour business terms from dbt, Snowflake, Unity Catalog, Jira and Confluence, the health of the context your AI uses, and an API your agents can query. The source readers are built, but not yet run against a live account.Ontology
- PeoplePaid seats nobody used, by team and service, from the seat export you upload. AI services used in the browser will appear once the browser extension is available. No individual person is named.Catalogue → People
- SavingsEvery saving found, in one list, with the method of calculation, whether it is measured, replayed or estimated, and whether it was realised.Spend → Savings
- EvidenceOne tamper-evident evidence ledger for every kind of record, readiness by article, sealed reports, incidents with their reporting deadlines, and AI literacy.Evidence
Optional, later
Additions with the Gateway
The console and the policy stay the same. The gateway pulls the policy and acts on each request sent through it.
The Gateway →- Enforce, the seventh area of coverage: a decision on each request sent through a gateway to allow, redact, route or refuse it
- A chained decision record for each request the gateway decides, for the evidence ledger and the evidence pack
- Emergency stops, checked by every gateway every 15 seconds in every mode
- Bills matched to recorded requests by provider, model and day
- Model recommendations from a replay of your own requests, once 100 requests are in the evaluation corpus
Data location for an EU organisation
- In Stockholm: the organisation's database, files and evidence, their backups, its keys, the sign-in identities of its people and the e-mail it sends.
- The model requests Saible makes for the organisation are processed inside the EU, in any of six EU regions, and never through a global profile.
- AWS's own activity log for Saible's EU account is kept in the US. It records which identity called which AWS service, when, from which IP address and with which parameters. For user administration, the parameters can include a person's sign-in name. The log holds no prompts, records, evidence or files.
- An organisation whose first admin uses a personal e-mail address (gmail.com and similar) is created in Ohio, in the US. Use a work address for an organisation in Stockholm. People invited into it are held there, whatever their address.
Steps
- 1Request access. Saible creates your organisation in its region: Stockholm for the EU, Ohio for the US.
- 2Open Settings → Providers and connect a billing feed.
- 3Add your commitments on Contracts.
- 4Register your applications in the AI catalogue.
- 5Read Readiness, then export the Evidence pack.
Control Plane topics
- Coverage01Control coverage, in seven areasThe controls each AI system has. The Control Plane covers six areas, and the gateway adds Enforce.
- Catalogue02Your AI catalogue, from what you already runAI systems are found on your platforms, in your code and at the gateway, then claimed by an owner who adds the purpose and the risk class. The browser extension is not yet available.
- Catalogue03What Saible discovers when you connect a sourceEach read of a source keeps what it found as a proposal. An organisation admin accepts or rejects each one, and nothing found is acted on before that.
- Spend04Spend feeds from your providersBilling data is read once a day and matched against what the gateway recorded.
- Control05A change, approved by a second personA change is proposed, replayed over past requests and approved. Every gateway then pulls it.
- Evidence06Evidence ledgers, checks and the evidence packEach record is chained to the one before it. If a record is changed, verification names it.
- Context07The context of each applicationRead from the application's own sources, confirmed by a person and signed. A change waits for a decision.
- Agents08Connect an agent to Saible over MCPAn application's agent reads its approved context with a signed receipt. A person's agent reads the console's API as that person and changes nothing.
- Evaluation09Candidate models, on your own promptsEvery request is decided by your policy and recorded. Figures are shown side by side, never as a verdict.
- Exposure10Configuration exposure of each AI platformRead-only, with the credentials you already gave Saible. Each finding comes with the script that fixes it. Built, but not yet run against a live account.