IntegrAuth Academy
Identity, security and practical AI — free, from zero. Byte-sized lessons, real diagrams, hands-on labs, zero fluff. Two programs, each with its own exam and certificate.
English is hard? Your browser can translate this page into Tamil, Telugu, Kannada, Malayalam or Hindi. Tech words stay in English so you learn the real names.
How?
- Chrome on phone: ⋮ menu → Translate
- Chrome on computer: right-click → Translate to …
- Edge: the translate icon in the address bar
Safe AI at work: data and privacy
Priya pastes an angry email from Maya, a customer, into her personal chatbot. It shows Maya's home address. Your goal: know where work data may go, and clean it first.
🧪 Interactive lab — enable JavaScript to play with this one.
- Everything you paste goes to the AI company. It can stay there after the reply.
- Training off does not mean deleted. Chats still sit in history and safety logs.
- Work data goes only into AI tools your company approved.
- Secrets never go into any AI tool, on any plan.
Personal plans and business plans
Free and personal plans often use your chats to train new models, unless you switch it off. Paid business plans usually do not, by default. Neither do paid API plans, which programs use. Some free API plans do. Business plans come with a contract your company controls. A personal account is outside that contract.
The paste rules
| What it is | Where it may go |
|---|---|
| 🌍 Public: published docs, open-source code | Any tool |
| 🏢 Internal: code, plans, meeting notes | Only AI tools your company approved |
| 🔐 Personal data: customer emails, HR records | Only a tool approved for that data. Otherwise, remove the details first. |
| ⛔ Secrets: passwords, API keys, tokens | Never, in any tool |
Clean it before you paste
Swap names for roles, like [customer]. This stand-in is a placeholder. Do the same with numbers. Cut a log to the lines that matter. Logs often hold a login token or a session cookie. A session cookie is a small saved ID that keeps you signed in. Delete them. The reply is often just as good. The AI does not need Maya's address.
A pasted key has left your control, even in an approved tool. Treat it as leaked. Replace it now. This is called rotating the key. See secrets management.
Now clean two work documents.
🧪 Interactive lab — enable JavaScript to play with this one.
Go deeper: where your text goes
A chat window feels private, like a notebook. It is not. Each message travels to the provider's servers (the computers it runs). Your text can end up in these places:
- The answer: the part you see.
- Your chat history: often kept until you delete it.
- Safety logs: providers often keep chats for a set time to catch misuse. This is a safety log. It happens even with training off.
- Human review: staff may read chats that safety systems flag.
- Training data: on some plans, chats help train future models.
- Legal hold: a court order can stop deletion.
Go deeper: plans, settings and connected apps
Business plans also keep logs for a set time, written in the contract. Some business customers can get nothing stored after the reply. This is zero data retention. As of October 2026, major providers say they do not train on paid API and business data by default. Some free API plans train on your data, like Google's free Gemini API. Terms change often, so check your plan and tier.
In India, a company must also protect its customers' personal data under the DPDP Act. A customer's details in an unapproved tool put that duty at risk. See Regulations & certifications.
On a personal plan, switch off training. Use a temporary chat. On most big products, it is not used for training and is deleted after a short time. Delete old chats. Don't connect apps like email or calendar unless the tool needs them.
A connected app can read and act with your access. Give it only the permissions it needs. These are called scopes.
Go deeper: who owns the output
In the US, only a human can be an author. This is human authorship. A prompt alone is usually not enough. Your own choices and edits can be protected. Other countries differ. You are responsible for what you publish, even if AI wrote the first draft. See Who owns AI output?
Follow your employer's, client's or school's rules on saying when AI helped. Under the EU AI Act, people must know when they talk to an AI. Deepfakes and other AI-made content must be labeled or marked. The Act's risk levels and dates are in AI rules at work.
Open your AI chat tool's settings. Do your chats train its models? Find temporary chats and how to delete your history.