Skip to content

Menu

AI Governance & Permissions Anyone can open the payroll file. Your AI will find it.

We set up your AI on accounts you own.An in-house AI team sets up your AI on accounts you own, with a record of who approved what.

This form needs JavaScript.

Reach us another way:

WhatsApphello@supertoasty.com

Tell us which AI tools your team uses

Our team will review your details and contact you about the next steps.

  • Your team keeps using AI
  • The accounts stay yours
  • A dedicated account manager

Banning ChatGPT

Ban ChatGPT on the laptops. Staff use their phones.

A ban only stops the AI use you can see.

In over 90% of companies, staff use personal AI tools for work. Only 40% have bought official AI subscriptions. MIT NANDA, reported by Fortune, 19 Aug 2025, checked 16 Nov 2026.

The fix: one business AI everyone may use, on accounts you own.

Staff on personal ChatGPT accounts today?

Get Started

Where it leaks

Forgotten shares and logins. AI can open every one.

Each is a share or login someone forgot. Closing it before AI goes live saves the clean-up.

She left in March. Her login didn’t.

Logins in the business’s name, so nothing leaves with a person.

  • “Read-only” that can still delete.

    OWASP lists this among the top AI risks.

    Check each AI’s rights in the app itself.

  • Apps staff let in years ago.

    Google’s admin console lists the apps that have read your company’s data.

    Go through the list before AI goes live.

  • The old agency is still an admin.

    Past suppliers can keep access to your CRM and ad accounts.

    Remove the ones you no longer use.

  • New AI tools nobody checked.

    IBM: only 34% of firms with AI rules regularly check for unapproved AI.

    Ask about new AI tools at each check-in.

    Cost of a Data Breach Report 2025, 30 Jul 2025, checked 16 Nov 2026.

97% of organisations with a breach of their AI models or apps had no AI access controls in place. IBM, Cost of a Data Breach Report 2025, 30 Jul 2025, checked 16 Nov 2026.

The order

Find, fence, then switch on.

Most of the work happens before AI is connected, by an in-house AI team working as part of yours.

A dedicated account manager. One person who knows your business and owns the work.

  1. 1

    Map how your team uses AI today

    Which tools, on which accounts. No blame.

  2. 2

    You choose what AI may see

    Where closing a folder would break daily work, hide it from AI instead.

    You decide
  3. 3

    Start from the tools you have

    One AI setup, on accounts you own.

    You approve
  4. 4

    Write down who may do what

    Each step and who owns it, with approvals routed to the right person.

    You sign off

After you press Submit

  1. 1We review your details.
  2. 2We talk it through with you.
  3. 3We propose a solution.

Staff AI rules

One page of AI rules your team will read.

An AI policy people follow fits on one page, with your own examples.

Client questions

Client asks about your AI? The answer is ready.

A record of who approved what, on accounts you own, is most of the answer.

Questions

Who sees what, and what AI learns.

Each answer links to the vendor’s own page.

What is AI governance and permissions?

AI governance and permissions sets who, and which AI, may see, use and change a business’s information: which AI tools staff use and on which accounts, which files and apps each AI can open, and who approves what it does. SuperToasty, an in-house AI team, sets it up on accounts the business owns, routes approvals to the right person and keeps a record of who approved what.

Who it suits

An owner-led business without its own security team, where staff already use ChatGPT, Claude, Copilot or Gemini, often on personal accounts. Also a supplier whose client, insurer or tender now asks in writing how it uses AI with their data.

Our staff use their own ChatGPT for work. What should we do?

Staff using personal ChatGPT for work is common: Microsoft and LinkedIn found that 78% of AI users bring their own AI tools to work (2024). A ban only moves it onto phones. List what is in use, without blame. A business account the company owns is the safer home for it. A personal account leaves when its owner does; a business account stays.

Will ChatGPT or Claude train on our company data?

ChatGPT’s business plans and Claude’s work plans do not train on a company’s chats by default, according to OpenAI and Anthropic. ChatGPT’s Free, Go, Plus and Pro plans do, unless the user opts out. Google says Workspace does not train on customer data without permission. Training is only half the risk: who owns the account, and who keeps it after someone leaves, matters as much.

Can Copilot or Gemini show staff files they shouldn’t see?

Microsoft 365 Copilot and Gemini can show a person any file that person’s account can already open, including one shared with the whole company years ago. Microsoft says Copilot “only surfaces organizational data to which individual users have at least view permissions”, and Google says the same of Gemini. The risk is old sharing, which AI makes easy to find. The fix is to close or fence off what is too wide before switching AI on.

What should an AI policy for staff include?

A staff AI policy works best on one page: which AI tools and accounts to use, what never goes in (card numbers, health records, passwords, a client’s confidential files), who to ask, and when the page is next reviewed.

A client asked how we use AI with their data. What do we tell them?

A client’s question about AI is answered with what is true in your business: which AI tools staff use and on which accounts, whether the vendor trains on the data, who can see what, and what is recorded. We keep a record of who approved what, on accounts you own, so the answer is ready. We are not lawyers or auditors, and we issue no certificate.

Is there an AI governance standard we should follow in Singapore?

Singapore’s IMDA published its Model AI Governance Framework for Agentic AI in January 2026. Its first area is to assess and bound the risks upfront, for example by “setting limits on agent autonomy, tools and data access”.

Every source linked in these answers was checked 16 Nov 2026.

A question we missed? Ask it in the form’s last box.

Get Started