1. How AI may assist our work
Paulioskos may use AI-assisted tools as aids for drafting, summarisation, content preparation, administrative efficiency, scripting support and similar low-risk productivity tasks. AI supports human work; it does not replace professional judgement, responsibility or appropriate technical review.
Outputs may be incomplete, inaccurate or unsuitable for a particular environment. A person reviews and, where necessary, tests or validates AI-assisted output before it is relied upon or shared.
2. Principles
- Human oversight. A person remains accountable for the task, considers the context and decides whether an output is appropriate.
- Data minimisation. We aim to submit only the information reasonably needed for the task and to remove unnecessary identifying, customer-specific or sensitive detail.
- Confidentiality. AI use does not change Paulioskos’ confidentiality expectations. Information remains subject to the same care, access and handling considerations that apply outside an AI-assisted workflow.
- Proportionality. The tool, information and proposed use should be proportionate to the sensitivity and consequence of the work. Particularly sensitive matters may be handled without AI assistance or through a more controlled method.
- Output validation. AI-assisted content is reviewed for accuracy, security, relevance, bias, unsupported claims and unintended disclosure. Code or technical guidance is tested and assessed before production use where appropriate.
3. Information normally excluded from prompts
We do not knowingly submit information such as:
- customer-specific confidential information where disclosure has not been appropriately considered;
- passwords, credentials, authentication tokens, private keys, recovery codes or other secrets;
- full sensitive configurations, unredacted logs or deep technical detail that could reasonably facilitate abuse if disseminated;
- unpublished vulnerability details, exploit paths or proofs of concept relating to a customer or Paulioskos environment;
- sensitive internal architecture, security-control placement, recovery locations or management routes;
- private IP schemes where their context or disclosure would be sensitive; or
- personal information or commercially sensitive material that is not reasonably necessary for the task.
Where technical assistance is appropriate, information should be reduced, sanitised, generalised or anonymised as far as reasonably practical. Examples may be reconstructed using placeholders, generic topology, synthetic data or the minimum isolated excerpt needed to explain the issue.
4. Professional judgement and customer matters
AI-assisted output is not treated as authoritative merely because it is fluent or plausible. Important decisions continue to require appropriate evidence, engineering judgement, security review, testing and customer context. Paulioskos remains responsible for the work it delivers.
Customers may raise concerns about AI-assisted handling of a particularly sensitive matter through the normal contact route. The appropriate approach can then be considered in the context of the information, engagement and available safeguards.
5. Scope and continuing development
This statement describes general practice. It does not claim that every tool is entirely local, that no external processing occurs, or that a particular product or safeguard applies to every use case. Tooling, provider terms, safeguards and business processes may change; Paulioskos will review its approach as they evolve.
This statement supplements rather than replaces contractual confidentiality commitments, the Website Privacy Notice, security processes or other controlled material in the Trust Centre.
6. Contacts
| Purpose | Route |
|---|---|
| Privacy questions and rights | [email protected] and the Website Privacy Notice |
| Security concerns | [email protected] |
| General or customer enquiries | Contact Paulioskos |
| Document questions | [email protected] |