At most one email a week. Unfollow anytime with one click.
Transcript
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Show transcript
Introduction
Hi, and welcome to the final episode of the Getting to Know Claude session. We have an account, we can hold a conversation, and we know how to choose a model. Now for the part you shouldn't skip: safety. In this episode we sort safe data from risky data, learn to anonymize data, check privacy settings, recognize Claude's limitations, and then write usage rules for a team.
Step 1: Safe data and risky data
Picture our data in three zones, like a traffic light. The green zone is safe to paste: a published price list, promo drafts, or Mr. Hendra's lesson material. The yellow zone needs anonymizing first: customer names and chats, or sales data per person. You can use the content, but replace their identity with a code or initials. The red zone should never be pasted into any AI: passwords, PINs, one-time codes, bank account or card numbers, ID card numbers, and confidential company documents without your manager's permission. Here's a simple test: if someone else read this message, would anyone be harmed? If the answer is yes, stop and anonymize it first. Why stay careful? First, your chat history is stored in your account. If your laptop is borrowed or someone else opens your account, every chat can be read, including what you wrote months ago. Second, any account can be hacked. Protect your Claude account with a strong password and don't share your login with lots of people. Third, rules apply: the policy of the company you work for, plus data protection laws, such as Indonesia's personal data protection law. For Dewi at Maju Bersama Co., this means she first needs to know whether her office allows internal documents to be processed with AI, and which tools are approved.
Step 2: Anonymizing data before pasting
Let's practice anonymizing data. Dimas receives a complaint about a hamper. The original message includes the customer's full name, phone number, and office address. Dimas replaces them: the name becomes Customer A, the phone number is removed, and the address is shortened to just the district name. The problem is still clear, but the identity is protected. Claude can still write an accurate, polite reply, because what it needs is the problem itself: two hampers and one missing greeting card, not who the person is. Once the reply is ready, Dimas greets the customer by name directly in WhatsApp. The real identity never enters the conversation with the AI. To make it clearer, here's a before and after table. The customer's full name becomes Customer A, or initials if you need to tell several people apart. The phone number is removed completely. Claude doesn't need it to write a reply. The full address is shortened to the district. That's enough to work out shipping or estimate delivery time. Bank account details are replaced with a general description, such as payment by bank transfer. Make anonymizing a habit before you paste. After a few times, this step takes less than a minute.
Step 3: Checking privacy settings
Now open the settings menu, privacy section. This is a simulated screen; option names and positions may differ in your account. The first option to check: whether your chats may be used to help improve the models. Choose what you're comfortable with and what your workplace policy allows. Second, managing your history. You can delete specific chats, for example ones that accidentally contain sensitive data. Third, exporting data. You can request a copy of your account data for your personal archive. There's no wrong choice here. What matters is that you choose consciously, instead of leaving the settings as they are without ever opening them. Note your choices so they're easy to review later. Privacy policies can be updated at any time. Read Anthropic's official privacy policy page regularly, especially before using Claude for office data. For office use, there are three extra points. First, ask about your office policy: which AI tools are approved and what data may be processed. Second, consider a plan for teams or enterprises. These plans usually have stricter data controls and member management. Check the details on the official page. Third, don't share one account among many people. Besides possibly breaking the terms of service, everyone's chat history gets mixed together and becomes hard to oversee. If you work alone, like Nadia the freelancer, just make sure your personal account is protected and client data is anonymized before you paste it. The principle is the same.
Step 4: Recognizing Claude's limitations
Now the second limitation: Claude can be wrong. Laras asks about this year's small business tax rate. A question like this involves rules that change often. A good answer from Claude actually admits its limits: it gives a general picture, then suggests Laras check the official tax website or a consultant. Laras takes a smart step: she asks which part most needs checking. You can use this question for any topic. Claude names three things to verify. Now Laras knows exactly what to ask the tax officer. Remember, for law, tax, health, and finance, Claude is a thinking partner, not a replacement for an expert. Final decisions are still based on official sources. There are three safeguard sentences you can add to your prompt. First: flag the parts you're not sure about. Claude will point out the weak spots in its answer. Second: explain the basis for your answer. That way we can judge whether the reasoning makes sense. Third: if you don't know, say so. This sentence helps reduce answers that sound confident but are made up. For example, Mr. Hendra adds a safeguard sentence when asking for a summary of history material. Claude flags two dates to check against the textbook, and one of them turns out to be wrong. For the latest information, such as market prices or news, use the web search feature when available. We'll learn it in the Research with Claude session.
Step 5: Safety rules for you and your team
Finally, Laras writes AI usage rules for Sunset Coffee and pins them near the admin desk. Rule one: anonymize customer names, phone numbers, and addresses. Two, never paste passwords, one-time codes, bank account numbers, or ID numbers. Three, check prices, dates, and rules against official sources before using them. Four, reread every reply before sending it to a customer. Five, one person, one account, with a strong password. Dimas has his own account. Six, delete chats that accidentally contain sensitive data. These six rules fit on one page, yet they protect the business and its customers. Feel free to copy them and adapt them for your own team. One principle sums it all up: Claude is the assistant, and you remain the one ultimately responsible. Protect data, check facts, and read results before using them. With this principle, you can use Claude with confidence, without excessive fear and without being careless. The Getting to Know Claude session is now complete. You have an account, know the interface, can hold useful conversations, choose a model, and use Claude safely.
Done!
We've sorted data into three zones, anonymized customer data, checked privacy settings, recognized Claude's limitations, and written safety rules for a team. The Getting to Know Claude session is done. In the next episode we start the Prompt Basics session, beginning with prompt structure: role, context, task, and format. See you there!
Practice exercises
Try these on your own after watching to lock in what you learned.
Anonymize one real customer conversation before asking Claude to draft a reply
Check your account's privacy settings and note the choices you made
Write five AI usage rules for your team or family business
Ask the class material
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
My notes
Notes are saved with the current video time and are only visible to you.
Questions & answers
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.
Your instructor
Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
53 video classes1,290 episodes
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
