You deleted the AI app. Did you actually disconnect it?
Deleting an AI app from a laptop does not always remove the permissions it was given elsewhere. Here is how to check Google, Microsoft, Slack, GitHub and the data the service may already hold.
Practical AI safety
Guides for safer AI use.
Clear, practical explanations of the AI safety, privacy and trust questions that matter at work. Every published guide is source-linked, reviewed and dated.
Latest guides
Before you let AI patch production: 7 checks
AI security tools are starting to find and fix vulnerabilities automatically. Before you let one change production systems, check its permissions, approval steps, logging, isolation and rollback plan.
20 AI safety checks a small business can do in 20 minutes
A fast, practical 20-point check for small businesses that want a clearer picture of how staff use AI, what it can access and what needs fixing next.
Your employee just connected an AI tool to work. Now what?
A calm, practical response for small businesses when a team member connects an AI tool to company email, files or other work systems without a clear approval process.
“Sign in with Google” is convenient. What did you just give away?
Signing in with Google or Microsoft is often just a login. Sometimes it is also a permission grant. Here is how to tell which one you just clicked.
That AI meeting bot just joined your call. Who invited it?
An AI notetaker appears in your meeting and nobody remembers inviting it. Here is how to work out who connected it, what it can capture, where the notes go, and whether it should stay.
How to use Figma with AI without opening up your whole workspace
A practical guide to using Figma with AI while limiting access to the files, prototypes and permissions the workflow actually needs.
Can AI leak your data without being hacked?
Sensitive information can be exposed through normal AI use without a traditional hack. This guide explains how prompts, connectors, retrieval and outputs can move data beyond its intended audience.
Too much power: why AI agents shouldn't have access to everything
A practical guide to excessive agency: why AI agents should have only the tools, permissions and autonomy they need.
Prompt injection: when AI reads instructions you never gave it
A plain-English guide to prompt injection, why hidden instructions in websites, emails and documents matter, and how permissions can limit the damage.
Can I paste this into AI? A practical guide for business information
A practical guide for deciding what information is lower-risk to share with AI, what needs checking first, and what should stay out.
How to let AI help with email without giving it your whole inbox
A practical small-business guide to using AI with email while keeping the connection limited to the messages and mailbox it actually needs.
What happens to your data after you send it to an AI?
A practical guide to what can happen to prompts, files and outputs after you send them to an AI service - including processing, retention, logging, human access, model improvement and subprocessors.
How to give AI access to Google Drive without giving it everything
A practical small-business guide to letting AI work with selected Google Drive folders without connecting an account that can see the whole company Drive.