AI privacy

AI Privacy Concerns at Work: What Should a Small Firm Worry About?

Updated November 16, 2026

Short answer: AI privacy is about what happens to the information you type or upload into an AI tool: who keeps it, for how long, who can read it, and whether it is used to train the model. For a small firm that holds client data, five AI privacy concerns matter most: staff using personal accounts, models trained on what goes in, chats kept on the vendor's servers, vendor staff who can read stored chats, and sharing features or bugs that expose chats. AI data privacy is decided mostly by the plan and its settings, not the brand name. So good AI privacy and data protection starts with an approved tool, a short written policy and settings chosen on purpose.

What is AI privacy?

AI privacy means keeping control of personal and confidential information when it goes into an AI tool, or comes back out of one. The U.S. government's AI risk framework lists data privacy as one of the main risks of generative AI, which it describes as "Impacts due to leakage and unauthorized use, disclosure, or de-anonymization of biometric, health, location, or other personally identifiable information or sensitive data" (NIST AI 600-1).

For a small firm, the practical test is simpler. Once a client's file goes into an AI tool, can you say where it is, who can see it, and how to delete it?

What are the main AI privacy concerns for a small firm?

  1. Personal accounts. The biggest concern is the account nobody approved. Cyberhaven, which sells data-loss-prevention software, reports that "32.3 percent of ChatGPT usage happens through personal accounts", and that "Personal accounts bypass SSO enforcement, centralized logging, enterprise retention policies, and controls related to data usage or model training" (Cyberhaven, February 2026). See Can staff use their personal ChatGPT accounts for work?
  2. Training on what goes in. On ChatGPT Free, Plus and Pro, "data sharing is enabled by default" until the user turns off "Improve the model for everyone" (OpenAI). NIST notes that models "may leak, generate, or correctly infer sensitive information about individuals" (NIST AI 600-1).
  3. Chats kept on the vendor's servers. "No training" is not the same as "not kept". Claude Team does not train on your content by default, and also says "we retain your chats and coding sessions in the product" (Anthropic). ChatGPT Business removes deleted chats "within 30 days, unless longer retention is required by law, or is reasonably necessary to protect our services or any third party from harm" (OpenAI).
  4. People at the vendor. OpenAI says business content can be seen by "authorized employees that require access for engineering support, investigating potential platform abuse, and legal compliance" (OpenAI). Read who and why for any tool you approve.
  5. Sharing features and bugs. In 2023 a ChatGPT bug "allowed some users to see titles from another active user's chat history" (OpenAI). In 2025 OpenAI removed an option that let shared chats show up in search engines, "to prevent users from unwittingly exposing sensitive information" (The Register). See What happens when client data leaks through ChatGPT?

How common are AI data privacy problems at work?

Common enough that most firms should assume some client data has already gone into an AI tool:

  • Company information going in. Cisco's 2024 Data Privacy Benchmark Study, drawn from "2,600 privacy and security professionals across 12 geographies", found that people had entered "employee information (45%) or non-public information about the company (48%)" into generative AI tools. The top concerns were "threats to an organization's legal and Intellectual Property rights (69%), and the risk of disclosure of information to the public or competitors (68%)" (Cisco, January 25, 2024).
  • Without permission. In a CybSafe and National Cybersecurity Alliance survey, 38% of employees said they "share sensitive work information with AI tools without their employer's permission" (Infosecurity Magazine, September 2024).
  • Breaches. In IBM's 2025 breach study, "One in five organizations reported a breach due to shadow AI" (IBM).

These surveys mostly cover larger employers. We found no survey limited to firms under 25 people.

Why AI privacy matters to your clients

Your clients chose you partly because they trust you with their information. In the same Cisco study, "94% of respondents said their customers would not buy from them if they did not adequately protect data." Many sectors also have written rules about the outside companies that handle client data. The sector risks page sets them out for insurance agencies, advisers, accounting firms and professional services.

How do you protect data privacy when using AI?

Choose the tool on purpose, write down what may go in, and set the tool up so it keeps as little as possible. In practice:

  1. Approve one tool on a plan your firm owns. Not a collection of personal accounts. What makes an AI tool safe to approve has the five questions to ask.
  2. Write a short AI policy. Say which tools are approved, what client data may go in, and what never does. Our AI acceptable use policy template is free to copy.
  3. Set it up on purpose. Turn training off where there is a choice, set the shortest retention that fits your records, add only current staff, and use single sign-on where the plan offers it. How do I stop employees pasting client data into ChatGPT? covers the settings.
  4. Put in only what the task needs. Leave out names and account numbers when the task works without them.
  5. Make it safe to report a slip. Staff who know they can report a mistake the same day give you the chance to delete it and check your duties.

Does a business AI plan solve AI privacy?

Partly. ChatGPT Enterprise and Claude Enterprise both say they do not train on your business data unless you opt in, and ChatGPT Business says "By default, we do not use your business data for training our models" (OpenAI). That answers the training concern. It does not answer the others: business plans still keep chats, and some vendor staff can still see them for set reasons.

A private AI workspace is another option. With Faradex, only your team can see your files, and your conversations live only in memory, with no backups: zero training, zero retention. Faradex does not monitor or block other AI tools, so it works alongside your policy, not instead of it. What makes an AI tool safe to approve compares it with ChatGPT Business, Claude Team and Microsoft 365 Copilot.

About Faradex. Faradex is a private AI workspace for small and mid-sized firms that handle confidential client work. Each firm gets its own Workspace on dedicated infrastructure in the United States. Files and conversations are held in memory and are gone when the Workspace is reset. AI model: Currently Opus via AWS Bedrock, more information is available on our website. AWS does not store prompts or responses, and model providers have no access to them. Pricing: from $120 a month, no seat minimum; see faradex.ai/pricing.