Internal guidelines · AI use

Using AI safely at LTL

How we protect student, teacher, partner and company information when we work with AI tools.

Version 1.3 · 27 October 2026 · Applies to all LTL colleagues

Go to the confirmation ↓
The short version

1The three rules

1
Confidential data stays in approved AIContracts, salaries, unpublished drafts, campaign plans, prices not yet announced, and any non-marketing information about students, teachers, partners, providers or colleagues is confidential. It may only be used in approved AI (Claude by default, Doubao Work Team or Enterprise in mainland China, or another tool approved by the AI Manager), and even there it must be handled carefully (see section 3).
2
Public data can be used more freelyPublic material such as marketing, brand and design content can be used in Claude, in other company-managed business or team AI tools, and in everyday tools that merely include AI features.
3
New assets in dedicated AI tools are created in approved AI onlyAny new asset, artifact or other material created in a dedicated or primary AI tool (such as Claude) is created only in a pre-approved AI. Creating public marketing visuals in Canva and similar everyday tools — which are not primarily AI tools — is fine under Rule 2.

2Which rule applies?

Type Examples Where AI is allowed
Confidential Contracts and agreements; salaries and pay; student information (applications, ID documents, payments, progress, feedback); information about teachers, partners, providers and colleagues; unpublished drafts, campaign plans and prices not yet announced; any other internal, non-marketing business information. Claude, Doubao Work Team or Enterprise (mainland China), or another AI approved by the AI Manager, using your company account, handled carefully.
Public Published marketing copy, brand and design material, public website content, social media posts. Claude, other company-managed business or team AI, or everyday tools with built-in AI features.
New assets Anything you create in a dedicated AI tool that did not exist before: documents, program outlines, tests, web pages, artifacts, images, code. Public creative work in Canva and similar everyday tools falls under Rule 2. Claude, Doubao Work Team or Enterprise (mainland China), or another pre-approved dedicated AI.

Rule of thumb: if you are not sure whether something is public, treat it as confidential. Unpublished drafts and prices not yet announced are confidential until they are released.

3Handling confidential data in approved AI

Claude, Doubao Work Team or Enterprise (in mainland China) and any other AI the AI Manager has approved are the approved places for confidential material, but “approved” does not mean “careless”. Please:

Projects in approved AI tools
Any project you create in a pre-approved AI tool (such as a Claude project) that contains uploaded documents, artifacts or other files must have AI Manager Alex Krasnov added to it. Add him when you create the project, before you upload anything. This lets us see what is stored where and keep it protected.

4Using AI tools other than Claude

5Training, ownership and content rules

6Accounts, devices and leaving LTL

7Quick do and don’t

Do Don’t
Use your company account on an approved AI for anything confidential.Paste contracts, salaries or student details into any unapproved AI tool.
Remove names and ID details when the task does not need them.Use a free or personal (consumer) AI chatbot account for non-public company information.
Create new documents, tests, pages and other assets in Claude or another pre-approved dedicated AI.Create new assets in an unapproved dedicated AI tool.
Review AI output before sending it on.Send AI output to students, partners or providers unchecked.
Add AI Manager Alex Krasnov to every project that contains uploaded files or artifacts.Keep a project with company files private to yourself.
Ask when you are unsure, before you start. Apply through the Special Approvals Application worksheet before using a new dedicated AI tool.Assume a tool is fine because a colleague uses it.
Select the no-training data option in any AI tool that holds company information.Leave company data open to AI training when the tool offers an opt-out.
Use AI agents and connectors to company systems only via approved team/business AI.Connect a personal consumer AI agent to company email, Drive or Mingdao; or leave company data in personal accounts after you leave.
Tell participants before you record or transcribe a meeting with AI.Enter passwords, API keys, or bank or card numbers into any AI tool, ever.

8If something goes wrong

If you paste confidential information into the wrong tool or account, share a link by mistake, send AI output to the wrong person, or an AI agent does something you did not expect, tell AI Manager Alex Krasnov the same day (alexander@ltl-school.com). Do not try to delete it quietly yourself first without reporting. Mistakes happen; speed matters more than blame. Reporting early gives us the best chance to limit the damage, for example by deleting the data or changing a password.

9Consequences

Breaches of these AI guidelines will be dealt with seriously. The rules are here to protect company information and the trust of our students and colleagues.

10Approvals, questions and updates

AI tools and their terms change quickly, so these guidelines will be reviewed at least every six months, and new colleagues are introduced to them when they join. To request a new AI tool or new AI permissions, apply through the Special Approvals Application worksheet. If you are unsure whether something is allowed, first check what you are already permitted to use. For other questions, ask AI Manager Alex Krasnov. If a translation differs from the English version, the English version applies.

Confirm that you have read and understood

Your signature

Thank you, your confirmation has been saved.

Download your copy (PDF)
Lex the Lion working on a laptop

Not sure? Just ask!

Using AI well is a team effort. If something is unclear, drop us a message. We are happy to help.