AI and Data Privacy Laws: What Businesses Need to Know

· DoubleX

Quick answer

Privacy laws like Brazil's LGPD apply to AI tools the same way they apply to any system that handles personal data: you need a purpose, limited access, and control. DoubleX Agentes does not train models on company content, but deciding what data agents touch stays with your business.

Managers often ask whether using AI at work creates a privacy problem. The short version: AI does not sit outside data protection law, and it does not need a separate rulebook to use it responsibly. This guide is a plain-language overview, not legal advice. For decisions about your own situation, talk to your legal or compliance team.

What does the LGPD say about data in AI tools?#

The LGPD (Lei Geral de Proteção de Dados) is Brazil's general data protection law, and it applies to any processing of personal data, including when an AI tool does the processing. If an agent reads a customer email, a CRM contact, or an HR spreadsheet, that is processing of personal data.

In practice, the ideas that matter most to a business are:

  • Purpose: use personal data for a clear, legitimate reason, such as answering a customer or preparing an invoice.
  • Necessity: give the tool access only to the data the task needs.
  • Security: protect the data from unauthorized access.
  • Transparency and accountability: be able to show what was done, with which data, and by whom.
  • Data subject rights: be able to find, correct, or delete a person's data when they ask.

Other laws, such as the GDPR in Europe or state laws in the US, share many of these ideas, but their details differ. Check which ones apply to your company.

What company data can an AI agent access?#

An AI agent can access only what you connect to it, and only with the permissions you give it. An agent is an AI assistant that works inside your systems, so its reach is defined by its connectors, documents, and permissions.

The usual sources of personal data are:

SourceExample of personal dataRisk to review
Email and calendarNames, addresses, message content, meeting attendeesAgent reads or sends messages you did not intend
CRMContacts, deals, support ticketsBulk edits to customer records
Spreadsheets and documentsPayroll, customer lists, contractsSensitive files shared too widely
DatabasesCustomer and transaction tablesQueries that return more than needed
Knowledge baseInternal notes and processesPersonal data pasted into notes

A useful exercise is to list, for each agent, which of these sources it needs. Anything it does not need should not be connected. For a deeper look, see Data Privacy and AI Agents: What Companies Need to Check.

What does DoubleX do, and what is up to your company?#

DoubleX Agentes handles the platform side of privacy and control, and your company remains responsible for how it uses the tool. Both halves matter.

What the platform does:

  • Company content is not used to train models, and requests go only to AI providers that do not use what they receive for training.
  • Each company has its own database, separate from other companies.
  • Passwords, tokens, and connection keys are encrypted with AES-256-GCM.
  • Roles (owner, administrator, member), teams, and sharing by person, team, or company.
  • For each agent, you choose which systems it uses and which actions it can run in each one. Actions that change data can be turned off, leaving read-only access.
  • Before sending, deleting, or changing something in your systems, the agent asks for confirmation and shows what it did step by step.
  • Every task keeps a full history: the request, each tool used, and the result.
  • The platform follows the LGPD, and its privacy policy is at https://doublex.ai/privacidade.
  • An optional zero retention mode uses only providers that store nothing after the response.

What stays with your company:

  • Deciding the legal basis and purpose for processing personal data.
  • Choosing which data and systems to connect, and to which agents.
  • Giving the right people the right roles and permissions.
  • Reviewing confirmations before approving actions.
  • Answering data subject requests and handling incidents.
  • Training your team on what not to paste into an AI tool.

No software makes a company compliant on its own. It gives you the controls, and you decide how to use them. These guides go further on each control: How to Control What an AI Agent Can Access and Do and Zero Data Retention in AI: What It Is and When to Use It.

What questions should you ask any AI vendor?#

Ask every AI vendor the same short list, and expect specific answers rather than general reassurance. You can copy these:

  • "Is our content used to train models, by you or by your AI providers?"
  • "Where is our data stored, and is it separated from other customers' data?"
  • "How are passwords, tokens, and keys protected?"
  • "Can we limit, per agent, which systems it uses and which actions it can take?"
  • "Does the tool ask for approval before sending, deleting, or editing, and is there a history of every action?"

Also ask whether the vendor offers a mode where providers keep nothing after answering, and how you can request deletion of data. To learn how to verify the training answer, read Does AI Train on Your Company Data? How to Check.

Where do you start safely?#

Start small, with read-only access and one low-risk routine, then widen the scope as you build trust. A practical path in DoubleX Agentes:

  1. Create a free account (no card needed) and run the diagnostic, a 5 to 10 minute conversation about your routine that shows where AI helps and where to begin.
  2. Turn the diagnostic into a team of agents, then open Agents and review each agent's instructions.
  3. In Connectors, connect only the systems each agent needs. Mark personal connections, such as your own email, as personal.
  4. Set permissions for each system. Leave actions that change data turned off at first, so the agent is read-only.
  5. Run a first request and check the task history in Tasks to see what the agent did.
  6. Add scheduled routines only after you trust the results. Routines run with the permissions of the person who created them.

For setup details, see the Quickstart.

Frequently asked questions#

No, the same rules apply: you need a clear purpose and a legal basis for processing personal data, whatever the tool. Your legal team should confirm the right basis for each use.

Is my company's data used to train AI models in DoubleX Agentes?#

No. Company content is not used to train models, and requests go only to AI providers that do not use what they receive for training.

Can an agent change or send things without my approval?#

Not unless you allow it. Before sending, deleting, or changing something in your systems, the agent asks for confirmation, and you can turn off actions that change data entirely.

Does DoubleX Agentes make my company LGPD compliant?#

No tool can do that alone. The platform follows the LGPD and gives you controls, but purpose, legal basis, and access decisions remain your company's responsibility.

How can I see what an agent did with my data?#

Every request becomes a task with a full history: who asked, which tools the agent used, and the result. You can review it at any time.

Related guides

More Security and privacy guides