Does Our Small Business Need an AI Use Policy? Texas and Colorado Rules in Plain English

By Dane Gray, 24uNet. Last reviewed November 9, 2026. This is general information, not legal advice.
Short answer: Yes. A one-page AI use policy fits almost any business whose staff use ChatGPT, Copilot or AI note-takers. Texas's AI law, in effect since January 1, 2026, targets intentional misuse. Colorado's law starts January 1, 2027 and covers AI-assisted decisions such as hiring or lending. The bigger reason is protecting client data.
Why write a policy at all?
If you haven't set rules for AI tools, your staff are setting them for you. One person pastes a client email into a free chatbot to tidy the wording. Another turns on a meeting recorder that stores every call somewhere nobody has reviewed.
Most of these choices are made to save time, not to cause harm. A short written policy gives people a clear answer to "can I use this?" and gives you something to point to when a client, an insurer or a regulator asks how your business handles AI.
What does Texas's AI law require of a small business?
Texas passed the Texas Responsible Artificial Intelligence Governance Act, known as TRAIGA (House Bill 149). It took effect January 1, 2026. It applies to anyone who promotes or conducts business in Texas, or develops or deploys an AI system there. Here is what the statute text says, in plain terms:
It targets intent. A person may not develop or deploy an AI system with the intent to incite or encourage self-harm, harm to others or criminal activity. Other prohibitions cover AI built with the sole intent to infringe constitutional rights, and AI deployed with the intent to unlawfully discriminate against a protected class. The law says a disparate impact alone isn't enough to show that intent.
Disclosure duties are narrow. The rule that requires telling people they're talking to an AI applies to governmental agencies, and to health care providers using AI in treatment.
The attorney general enforces it. There's no private right of action. Before suing, the attorney general must send a written notice and wait 60 days. A business that cures the violation in that window, and sends a written statement describing the fix and the policy changes made to prevent a repeat, avoids the lawsuit.
Penalties are set by statute. For a violation a court finds curable, $10,000 to $12,000 each. For one found uncurable, $80,000 to $200,000 each. Continued violations add $2,000 to $40,000 per day.
Two details matter for small businesses. First, TRAIGA doesn't contain a general requirement for private businesses to adopt a written AI policy. Second, it says a business may not be found liable if it finds a violation through its own internal review and substantially complies with the most recent NIST AI Risk Management Framework: Generative AI Profile, or another recognized framework. A written policy built on a framework like NIST's is the practical way to show that review exists.
What does Colorado's AI law cover?
Colorado signed Senate Bill 26-189 on May 14, 2026. It replaces the original Colorado AI Act and applies to consequential decisions made on or after January 1, 2027.
The scope is narrower than the name suggests. The law covers "automated decision-making technology" that is used to materially influence a consequential decision. Consequential decisions are those that relate to access to, eligibility for, selection for, or compensation for one of seven areas:
Education
Employment
Residential real estate (lease or purchase)
Financial or lending services
Insurance
Health care
Essential government services
A "deployer" is a person doing business in Colorado that deploys a covered tool. The definition doesn't mention business size, so being small doesn't remove you from the law if you use such a tool. Employees and Colorado-resident job applicants count as consumers.
If you are a deployer, the law's main duties are notice and recourse. Before using the tool, you tell consumers it's in use. After an adverse outcome, you give the person a plain-language description within 30 days. On request, you offer a way to correct the personal data used and a chance for meaningful human review, to the extent commercially reasonable. Records are kept for at least three years. The attorney general enforces the law and there's no private right of action. The attorney general must also adopt rules on the adverse-outcome disclosures by January 1, 2027, so details may still shift.
The law lists what it does not treat as a consequential decision, including tools used only to summarize or organize information for human review without producing a score or ranking, spreadsheets that need manual analysis, cybersecurity and fraud-prevention technology, and routine tasks such as scheduling, marketing or customer-service triage.
In everyday terms: using AI to draft emails or summarize meetings is probably outside this law. Using AI to rank résumés, score applicants or screen tenants for Colorado residents is the kind of use to look at carefully with your attorney.
What belongs in a one-page AI use policy?
Keep it short enough that people read it. Eight lines cover most small offices.
Approved tools. Name the AI tools staff may use, and say they must sign in with their work account.
Data rules. Never paste client personal information, account numbers, passwords or unreleased documents into a tool that isn't on the approved list.
Human review. A person checks anything AI produces before it goes to a client or into a record.
No automated decisions about people. Hiring, firing, credit, tenant, patient or benefit decisions are not made or ranked by AI without a person reviewing the result. Ask before you start.
Meeting recorders and note-takers. Tell participants before recording or transcribing, and use only approved tools.
Client-facing AI. If a chatbot or AI assistant talks to customers, say so clearly.
Report problems. If someone pastes the wrong thing into the wrong tool, they tell IT right away, with no blame.
Review twice a year. Tools and rules change quickly.
Which settings back up the policy?
A policy only works if the settings agree with it. For businesses on Microsoft 365, a few facts help.
Microsoft states that Microsoft 365 Copilot and Copilot Chat, used with a work account, respect your identity model and permissions, inherit your sensitivity labels, apply your retention policies, support audit of interactions and follow your administrative settings. Microsoft also states that prompts, responses and data accessed through Microsoft Graph aren't used to train its foundation models. The details depend on your subscription plan.
That cuts both ways. Copilot can only show what the signed-in person is already allowed to open, so years of loose SharePoint and OneDrive sharing become instant answers. Cleaning up permissions comes before turning Copilot on. Our guide on what Microsoft 365 Copilot can see walks through that cleanup, and prompt injection through external email explains another risk worth knowing.
Your business buys and owns its Microsoft 365 and Copilot licensing. What we do is help you set it up and secure it.
Where does NIST fit?
NIST publishes a free Generative AI Profile that companions its AI Risk Management Framework. It's voluntary and not tied to any one industry. A small business doesn't need all of it. Use it as a checklist for your policy: who owns AI decisions, which tools are in use, which data they touch, and how you'll learn about problems.
How 24uNet helps
We help with the practical parts: finding which AI tools your staff already use, setting up approved tools with work accounts, tightening Microsoft 365 permissions and sign-in rules, and training staff on the policy in plain language. The legal reading belongs to your attorney. We bring the technical side so the policy matches what your systems do. If you want this as a defined piece of work, a Service Authorization can list it. If you'd rather have it as part of everyday IT and security, see BusinessCare. We work with businesses in Denver and Bryan–College Station.
Frequently asked questions
Does Texas require my business to have a written AI policy?
No. TRAIGA doesn't contain a general requirement for private businesses to adopt a written AI policy. A policy is still a practical way to show you review your AI use. If the attorney general sends a notice of violation, a cure statement must describe the changes made to internal policies.
Does Colorado's AI law apply to a small business?
It can. The law applies to a person doing business in Colorado that deploys automated decision-making technology that materially influences a consequential decision, such as hiring or lending. The text doesn't set a size threshold, so size alone doesn't exempt you. Using AI only to draft or summarize isn't the target.
Can employees put client information into ChatGPT?
Not unless your business has approved the tool and its data terms for that use. Free personal accounts shouldn't receive client information. Decide which tools are approved, require work accounts and put the rule in writing.
When do Colorado's AI rules start?
The law applies to consequential decisions made on or after January 1, 2027. Texas's AI law has applied since January 1, 2026.
What is the NIST Generative AI Profile?
It's a free, voluntary companion to NIST's AI Risk Management Framework that helps organizations manage risks from generative AI tools. Small businesses can use it as a checklist when writing an AI policy.
Want your AI policy to match your settings?
We'll walk through how your business uses Microsoft 365, what staff are already using and which gaps to close first. For ongoing Microsoft 365 help, see BusinessCare. For a defined policy and settings project, start with a Service Authorization. Denver (303) 468-5515 · College Station (979) 256-5100.
Sources (verified November 9, 2026)
Texas HB 149 (TRAIGA), enrolled text: effective January 1, 2026; Sec. 551.002 applicability; Sec. 552.051 disclosure (governmental agencies and health care); Sec. 552.052-.056 prohibitions (intent-based); Sec. 552.101 AG exclusive enforcement, no private right of action; Sec. 552.104 notice and 60-day cure; Sec. 552.105 penalties and NIST-framework provision: https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.htm
Texas Business & Commerce Code Chapter 552 (codified): https://tcss.legis.texas.gov/resources/BC/pdf/BC.552.pdf
Texas Attorney General, Consumer AI Rights page (TRAIGA effective January 1, 2026; enforcement and penalty ranges): https://www.texasattorneygeneral.gov/consumer-protection/file-consumer-complaint/consumer-ai-rights
Colorado General Assembly, SB26-189 bill page (signed May 14, 2026; summary of duties): https://www.leg.colorado.gov/bills/SB26-189
Colorado SB 26-189, signed act text (definitions of consequential decision, covered domain, deployer; exclusions; effective date January 1, 2027; applies to decisions on or after that date): https://www.statebillinfo.com/bills/bills/26/SB26_189_signed.PDF
Colorado Attorney General, ADMT rulemaking page: https://coag.gov/ai/
Finnegan, "Colorado Replaces Landmark AI Act: An Overview of the New SB 26-189 Framework," May 26, 2026 (notice duties, 30-day adverse outcome notice, three-year records, AG-only enforcement, 60-day cure): https://www.finnegan.com/en/insights/articles/colorado-replaces-landmark-ai-act-an-overview-of-the-new-sb-26-189-framework.html
Microsoft Learn, "Enterprise data protection in Microsoft Copilot and Microsoft Copilot Chat" (permissions, sensitivity labels, retention, audit; prompts and responses not used to train foundation models; controls vary by plan): https://learn.microsoft.com/en-us/copilot/plan/enterprise-data-protection
NIST, "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile" (NIST AI 600-1): https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence
Photo: Pexels 16094041, "Man Using Laptop wit Chat GPT," Matheus Bertelli. Pexels License.




Comments