VECTORIA · Legal centre
AI and professional judgement
Vector helps with your work. You decide what to share, supervise actions and validate the result.
Edition dated 2 October 20262026-10-02-subscriptions-ai-v3
Applies to new subscriptions accepting this version.
What it can do
The agent interprets instructions and uses available product tools to inspect information, perform operations or prepare results. Its scope depends on the version, permissions and functions actually enabled. A persuasive explanation does not prove an action occurred or a calculation complies with a standard.
Check project changes, notices and tool results. For important tasks, keep a copy and review units, sources, geometry, quantities, assumptions and regulatory profiles. Understand the scope of any approval or review requested by the interface.
Data transmitted
Vector receives your instruction and necessary task context through Vectoria services and TensorX. This may include document fragments, names, properties or technical results. The privacy policy distinguishes Civil, Cost and Viewer. Opening a local file does not mean sending the entire file to the agent.
Do not include passwords, keys, payment card data, special-category data, criminal records or children’s data. Minimise or anonymise unnecessary personal or confidential information. Make sure you have permission to share project data.
MCP connection for external agents
Plans that include this connection allow compatible MCP agents. You choose and authorise the external client. Its account, terms, charges and data processing are separate from your Vectoria subscription. Restrict access to the tools and projects needed and disconnect when the work ends.
Vectoria Civil lets you connect your own desktop app, such as ChatGPT (by OpenAI, which includes Codex) or Claude Desktop and Claude Code (by Anthropic). The connection uses your account and subscription with that provider; Vectoria does not receive or store the credentials for those accounts. What the assistant uses from the project is sent to its provider and processed under your terms with it. This does not include an external provider in the price or guarantee its availability. The connection notice identifies where data is sent. ChatGPT and OpenAI are trademarks of OpenAI; Claude and Anthropic are trademarks of Anthropic. They are named only to identify the app you connect and do not imply endorsement.
Balance, limits and pending operations
Managed AI uses the credits included in your selected plan. The allocation is shown before purchase and in your account, under the subscription terms. A request may reserve credits and consume them when executed; stopping it does not undo usage already incurred. After an incident, a reservation may remain pending until its outcome is reconciled, avoiding accidental duplicate execution.
There are no automatic overage charges. Exhausted allowance stops new managed AI operations. Manual tools keep their product permissions. Image generation, when enabled, requires transmission confirmation and may use a different provider identified before generation.
Reviewing the result
AI can make mistakes, omit information, confuse references or suggest unsuitable actions. Instructions embedded in external documents are not necessarily trustworthy. Do not treat a response as the sole evidence of technical compliance, safety, quantity, price or feasibility.
Before delivery, compare results with original sources and reproducible calculations. Check the standard and version, geographical scope, units and relationship to the project. A competent professional decides whether a result is suitable for administrative, construction, contractual or financial use. Report discrepancies to [email protected] with necessary context while avoiding unnecessary sensitive data.