Choose one task before choosing an AI tool. What information would it need? What would a mistake cost? Who checks the result, and how would the team finish the work without it?

• Customer, employee, health, financial, legal, or credential data pasted into an unapproved tool.
• An answer published without checking the source.
• Automation sending a promise or decision to a customer.
• A paid AI feature added before a real task is named.

• Choose one reversible, non-sensitive task.
• Define the correct output and who reviews it.
• Test with fake or public information.
• Keep, change, or stop based on measured time and error, not hype.

Not into a personal account. There, chats stay in that person’s own history, outside the business’s control. If staff use one for work anyway, open Settings, then Data controls, and turn off Improve the model for everyone. OpenAI says new chats then won’t be used to train its models.

Two catches. OpenAI says a thumbs up or down on a reply can let that whole chat be used for training. That holds even after you opt out. And opting out does not delete saved chats.

OpenAI says it does not use content from ChatGPT Business or Enterprise workspaces to train its models by default. Even there, keep passwords, card numbers, and health details out.

Then give staff a three-line rule for a busy Saturday. What never goes in. Who checks the output. Who approves anything a customer will see.

Maybe not. Plans change, so check what your current plan includes before you buy anything new.

Google says the Gemini app is now a core service for most Workspace editions. It says your content is not human reviewed or used to train AI models outside your organization without permission. That promise comes from your organization’s Workspace agreement, so staff should sign in with the work account.

Microsoft says prompts and responses in Microsoft Copilot for your organization, formerly Microsoft 365 Copilot, aren’t used to train the large models it runs on. Use the work sign-in there, too.

The bigger risk is your own sharing settings. Both companies say their assistant works with files a person already has permission to open. If everyone can open the payroll folder, the assistant can pull from it for everyone. Do a permission review first.

Slow down. NYC’s Local Law 144 covers automated employment decision tools. That is the law’s term for certain software that helps make hiring or promotion decisions. Employers can’t use one unless it had a bias audit within the past year. A summary of the audit must be public. Candidates or employees must get the required notices. The city’s Department of Consumer and Worker Protection has enforced it since July 5, 2023.

If a hiring app offers AI ranking, ask the vendor for its bias audit before you turn it on. The city’s FAQ says the employer, not the vendor, is responsible for making sure the audit was done. Read the city’s AEDT page and FAQ, too. This summary reflects the city’s pages as of October 2026. It is not legal advice. Talk to an employment lawyer if you are unsure.

No. The FTC’s rule on fake reviews, in effect since October 2024, covers reviews from people who don’t exist, “such as AI-generated fake reviews.” Businesses can’t create or sell them. The rule also bans rewards that depend on a review being positive or negative. This is a summary, not legal advice.

Using AI to draft a reply to a real review is different. Read the draft, fix it, and post it in your own words.

We can help test a low-risk use for AI with the tools you have. People still check the work and make the decisions.

Get a plain-English second opinion

Product features and rules can change. Use the linked source as the current word.

• Google Search: Helpful, reliable, people-first content — Official content-quality guidance.
• OpenAI Help Center: Data controls in ChatGPT — Model-training setting, feedback exception, and default for Business and Enterprise workspaces. Checked October 19, 2026.
• Google Workspace: Generative AI in Google Workspace Privacy Hub — How Gemini in Workspace handles business content, training, and file permissions. Checked October 19, 2026.
• Microsoft Learn: Data, Privacy, and Security for Microsoft Copilot — Copilot training commitments, renaming from Microsoft 365 Copilot, and permission-based access. Checked October 19, 2026.
• NYC Department of Consumer and Worker Protection: Automated Employment Decision Tools (AEDT) — Local Law 144 bias-audit and notice rules for AI hiring tools. Checked October 19, 2026.
• FTC: Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials (Aug. 14, 2024) — Rule covering AI-generated fake reviews and sentiment-conditioned incentives. Checked October 19, 2026.