Aury, autonomy, and the limits of both

Aury is the intelligence inside Aurygine. He reads your company's actual state — decisions, evidence, artifacts, prior outcomes — before he answers, which is what separates him from a general-purpose chatbot answering from nothing.

He is software. He is not a person, not an employee, not a licensed professional, and not a legal representative of your company.

He can be confidently wrong. Language models produce fluent text whether or not the underlying claim is correct, and fluency is not evidence. Aury is built to prefer your recorded evidence over invention, but he can still misread, over-generalise, or fill a gap he should have flagged.

Treat everything he produces as a draft by a capable colleague who has not checked their sources: useful, fast, and requiring your judgement before it becomes a decision.

Nothing Aury produces is legal, tax, accounting, financial, medical, or regulatory advice, however specific it sounds. Company formation, contracts, tax treatment, employment, licensing and compliance all need a qualified human in your jurisdiction before you act. Aury will tell you when something is in that territory — but the obligation to check is yours.

Missions you have explicitly approved continue on our servers. Closing the tab, shutting the machine or losing connectivity does not stop approved work — that is the point of it.

Nothing runs that you did not approve. Aury proposes; you decide. A mission that has never been approved sits waiting, indefinitely and silently.

Aury never does any of these on his own initiative:

• Sending anything to a person outside your company
• Publishing anything publicly, under your name or your company's
• Spending money, or committing you to spending money
• Entering into an agreement on your behalf
• Deleting or overwriting company data
• Connecting or disconnecting an external account

Each of these passes an approval gate that shows you the exact action before it happens.

Autonomy is a setting, per company, and it starts at the most cautious level. Raising it widens what Aury may do between check-ins; it never removes the approval gates listed above.

Changing the level is recorded as an explicit authorisation, with the level, the date and who set it. You can see that record and you can lower the level at any time.

Control Tower has a pause switch that halts all running and queued work for a company. It takes effect on the server, not just in your browser, so it works even if you close the page immediately after. There is no confirmation maze and no cooling-off period on stopping — a stop is a stop.

When you connect an email, social or other account, Aury can act through it only within the scope you granted and only after approval. Access tokens are stored encrypted and are never shown back to you or to anyone else. Disconnecting an account in Control Tower revokes that access immediately.

Your prompts and the company context needed to answer them are sent to model providers acting as processors on our behalf. Your company content is not sold, and it is not handed over for third-party model training. If that boundary matters to your business, read the Privacy Policy before putting sensitive material in.

You are, once you act on it. Aurygine is responsible for the service working as described — that missions run, that approvals are honoured, that the pause switch stops things. We are not responsible for a business decision you took because Aury suggested it. That division is deliberate, and it is the same one that applies to any tool that gives advice.

Report it from the Help page inside the product, or write to support@aurygine.com. A wrong answer with the real context attached is the most useful thing you can send us.