Drafts, proposes, never files.
Aria answers with the article beside it and prepares the work ahead. Every change waits for a person unless an administrator lets low-stakes changes apply.
Answers, prepares, proposes — and waits for you.
Six things Aria does in a workspace. Live is in the product; coming soon is built and switched off until it is ready.
Answers with the article beside it
Law questions are answered from a verified registry; workspace questions from only what you can see and Settings allows.
Work waiting on the first day
After setup: your suppliers, three policy drafts, the AI-literacy course and up to 6 risks for each classified system.
Supplier fact sheets, from their pages
Aria reads the supplier's homepage and up to 5 linked pages; you approve the fact sheet and record the assessment.
14 kinds of proposal, one Inbox
Every change Aria would make lands in the Inbox, with what it read and the article behind it.
Only what an administrator allows
Approve each is the default; Auto-apply covers low-stakes changes on minimal-risk systems, each logged as Aria's.
A daily look at what changed
Once a day, Aria flags expiring evidence and due reviews: at most 10 proposals, none applied, plus a digest email.
Proposes, cites, waits for a person.
Both lists are product rules: in the agent test suite, one write applied without approval fails it.
A first pass at a policy, a FRIA or an incident notification, from your register, for a person to edit.
Which obligations a system owes, and what each one also counts toward.
Every answer names its provision, checked against the registry; one that does not resolve is flagged.
Give it a goal, and each write lands in the Inbox as a proposal with its reasoning.
Incident notifications, EU database registration and dossier submissions stay manual; Aria will not even propose sending one.
Approve each is the default; classification and changes to a high-risk system always need a named person's sign-off.
Aria is decision support, not legal advice: confirm with counsel, and the product records who confirmed.
Conversations are not used to train the model, and Aria does not browse the web mid-conversation.
Today · Ask · Run · Inbox · Activity · Settings
Today opens first: what Aria prepared, what it applied under Auto-apply and what waits on you.
Every proposal in one place.
Whatever produced it, a proposal waits here with what Aria read and proposes, until someone applies or dismisses it.
- Apply writes the record and the audit line together
- Where the floor says sign-off, the button reads Review & sign
- Filter by source: runs, chat, signals
An answer, with its article.
Ask in plain language or against a document you attach; the top-bar drawer knows the page you are on.
- Citations always on: the toggle to drop them was removed
- Attach PDF, DOCX, TXT, MD or CSV
- Conversation history is off by default
As a deployer of a high-risk system, 13 obligations apply. The core of Art. 26:
- Art. 26(1)Use system per provider's instructions for use
- Art. 26(2)Assign human oversight person(s)
- Art. 26(5)Monitor operation; suspend & notify on serious risk
What it reads, how far it goes.
Knowledge follows Frameworks: enable ISO/IEC 42001 and its corpus grounds Aria, with no second list to sync.
- Team names and document titles stay off by default
- Anonymise system names before anything reaches the model
- Autonomy per capability, with the policy floor on each row
Give Aria a goal.
“Review my unclassified systems and propose classifications.” Aria works inside the run’s scope and sends what it proposes to the Inbox.
What Aria did, with outcomes.
Every run and proposal, with an outcome an auditor can read; a run that produced nothing says so.
Two settings. One floor that cannot be lowered.
Approve each queues every change for a person; Auto-apply applies low-stakes changes and logs them. Neither goes below the floor.
The floor is the Act’s: a competent person oversees a high-risk system, and the provider, not software, reports a serious incident.
Auto-apply never acts beyond the asker’s role, and background work (the first run, the risk map, supplier research) is never applied.
Ask Aria on our public pages, without an account.
Coming to the Help Centre and the trust page: answers from public content only, with citations. It never sees a workspace.
- No account and no history kept: the browser holds the thread
- Up to 20 questions a day per visitor, paid for by Veritome
- The same EU-hosted model and citation check as the product
Take measures to support the development of AI literacy of staff and others dealing with AI systems on your behalf, taking into account their knowledge, experience, education and training, the context of use and the persons on whom the systems are used.
Free in the Academy: General Staff AI Literacy
An AI system, documented the way yours will be.
Aria is an AI system, so we document it the way we ask you to document yours.
Configuration, not training; Mistral holds the model-provider duties, Veritome the system-provider ones.
Your prompt, attachments, the context you leave on and retrieved passages, under a data-processing agreement, not used for training.
Latest published run, Eval 12 on 02.09.2026: 20/20 answers, 0 forbidden outcomes; a sample of every suite is published.
Straight answers about Aria.
What does Aria actually do inside Veritome?
Aria drafts, maps and cites from your register, prepares a new workspace from your setup answers and researches suppliers from their own pages. Any change lands in its Inbox for a person to apply or dismiss.
Can I ask Aria something without an account?
Soon. Aria is coming to the Help Centre and the trust page for visitors: no account, public content only, the same citation check, up to 20 questions a day. It never sees a workspace.
Can Aria change something without me?
Not by default. Approve each queues every change for a person. Switch a capability to Auto-apply and only low-stakes changes apply, each logged. Classification and anything on a high-risk system always need sign-off, and regulator-facing actions are never automated.
Which model does Aria run on, and where?
Mistral Small, served by Mistral AI from Paris under a data-processing agreement; the application and its database run on Hetzner in Nuremberg. Prompts are not used to train the model. The transparency page and sub-processor register have the rest.
What does Aria know?
The EU AI Act and its recitals, Veritome's regulatory registry, official guidance, the record templates and the corpus of every framework you have enabled. The EU AI Act is always on.
Is Aria legal advice?
No. Aria is informational decision support. Every substantive answer asks you to confirm compliance decisions with qualified counsel, and your audit trail carries the person who confirmed, not Aria's suggestion.
Is Aria itself governed under the EU AI Act?
Yes. Aria is an AI system, so it carries the Art. 50 disclosure on every screen and publishes its own AI System ID card, the same card the product asks you to publish for yours.
More in the help centre: Ask Aria — what the assistant can and can’t do. Aria’s legal notice is the AI Transparency Notice.
Give Aria a goal. Approve what comes back.
Register a system free and ask Aria what it owes. Every proposal waits for you.