Independent AI consulting & implementation, now taking on new projectsStart a conversation
Security & data

A careful approach to security and data

AI systems are only useful if they can reach your information and act in your tools. That is exactly why they need careful limits. This page explains how those limits are set, in plain terms.

Straight up: Georgy Consulting is a one-person practice and does not hold certifications such as SOC 2 or ISO 27001. Instead, the work is designed to fit inside your security and compliance framework, uses services you approve, and leaves you in control of every account and credential.

Architecture

Built inside your environment

Where possible, every part of the system lives in accounts your business owns. The one thing that usually leaves is the text a model needs to read, and only the part the task requires.

Your environment · your accounts
Your systemsCRM · ERP · drives · email
Scoped credentialsseparate, revocable
Agent or workflowruns in your account
Approvalsfor consequential actions
Audit loginputs, outputs, actions
Test setyour real cases
AI model providerbusiness or API terms, chosen and approved by you
Diagram: the agent, your systems, credentials, approvals, audit log and test set all sit inside your own environment and accounts. Only the data a task needs is sent to an AI model provider that you have approved.
Principles

Applied to every engagement

You own the accounts

Systems are built and run in accounts registered to your business: your cloud, your automation platform, your AI provider account.

Ownership001

Least privilege

Each system gets only the permissions its task needs: read-only where reading is enough, limited to specific folders, mailboxes or records.

Access002

Data minimization

Only the information a task needs is sent to an AI model. Where practical, personal or sensitive details are removed or masked first.

Data003

Human approval

Actions with real consequences, like sending to customers, moving money or changing important records, can require a person's approval.

Oversight004

Logging

Inputs, outputs and actions recorded at a level agreed with you, while keeping logs free of unnecessary sensitive data.

Audit005

Testing first

Systems are tested against real examples, including awkward and adversarial ones, before they're used for real work.

Quality006
AI model providers

Where your data goes when AI processes it

Most AI systems send text or documents to a model provider for processing. Which provider, under which terms, and in which region are decisions you should make knowingly.

  1. Business terms, not consumer apps

    Business data goes through providers' business or API offerings, whose terms on data use and retention differ from free consumer apps.

  2. Terms reviewed with you

    Before anything is built, you'll see how the chosen provider says it handles data (retention, training use, where processing happens) so your team can approve it. Providers change their terms, so this is checked for each project.

  3. Options for stricter requirements

    Where needed, models can be accessed through your existing cloud provider, in a specific region, or, for some tasks, run on infrastructure you control.

  4. The right model for each task

    Not every step needs the most powerful model. Matching the model to the task can reduce both cost and how much data leaves your systems.

During an engagement

How access to your systems is handled

  • Access is requested in writing, scoped to what the work needs, and granted by you
  • Work starts in test or sandbox environments wherever your tools provide them
  • Credentials used by AI systems are separate from personal logins, kept in a secrets manager, and never written into code or documents
  • Personal access is removed at the end, and you receive a list of every credential the system uses
  • NDAs and data processing agreements are signed on request
Access register · example
SystemScopeStatus
Accounting (test)Read onlyactive
Accounts inbox1 mailboxactive
Shared drive2 foldersactive
Consultant loginBuild onlyremoved
AI-specific risks

Risks that come with AI, and how designs account for them

No design removes these risks completely. The aim is to reduce them to a level you're comfortable with, and to make problems visible when they happen.

AI-specific risks and their mitigations
RiskWhat it looks likeHow designs respond
Wrong or invented answersConfident output that is incorrectAnswers grounded in your documents with citations, outputs checked against your data, human review where accuracy matters
Manipulated inputsText in an email or document trying to steer the system (“prompt injection”)Tight limits on what the system may do, instructions kept separate from content, approval for sensitive actions
Data leaking between usersAn assistant revealing something a person couldn't otherwise seeYour existing permissions respected rather than pooling everything together
Behavior changing over timeModel updates or new kinds of input changing resultsA saved test set re-run after changes, and monitoring of real use
Compliance

Working within your obligations

If you work under rules such as GDPR, HIPAA, financial services regulation or professional confidentiality duties, those requirements shape the design from the first conversation: which data may be used, which providers are acceptable, what must be logged and what needs human review.

Georgy Consulting is not a law firm and doesn't give legal advice. For regulated data, your legal or compliance advisers should confirm the approach meets your obligations, and the design documentation is written to make that review straightforward.

Questions about a specific project? Email info@georgyconsulting.ai.

Next step

Have a security questionnaire?

Send it over. You'll get straight answers, including where the answer is “that's handled by your provider” or “not yet”.