217. AI Legal Risks for Small Business Owners: Disclosures, Client Data & ChatGPT Evidence

Overview

The speaker argues that while small businesses initially treated AI as a convenient tool, it has rapidly become a real legal risk—especially for creatives (photographers, designers, copywriters, etc.). AI outputs can become evidence in lawsuits and can trigger copyright, privacy, and consumer-protection issues.

---

1) AI is creating real-world legal exposure for small businesses

• The core claim is that AI use is no longer “casual” or “private.” Courts, the FTC, and multiple state attorneys general are actively addressing AI-related deception and misuse.
• The speaker frames AI as creating both:
  • Liability, and
  • legal evidence—meaning what a business types or submits can later be used against it.

---

2) Photographers: a featured 2026 case shows how AI portfolios can become fraud and copyright violations

The speaker describes a major photography/wedding industry scandal (starting with a Reddit post and spreading through online accountability efforts) involving a photographer who allegedly used an alias (“Bonfage”).

Allegations described include:

• Real photographers’ work was allegedly lifted, converted into AI-enhanced content, and presented as the defendant’s own portfolio.
• Some “wedding photos” in the portfolio allegedly were fake—AI-generated and/or not from real weddings.
• The portfolio allegedly relied on images from associates as if they were the photographer’s own work.

Detection and purported inconsistencies:

• Detection tools allegedly flagged many images as AI-edited/generated.
• The speaker points to alleged inconsistencies (e.g., venue details not matching real-world photos).

Potential consequences emphasized:

• Copyright infringement (copying/uploading protected work into AI tools and re-presenting it as one’s own).
• False advertising / deceptive practices when clients pay based on a fabricated portfolio.
• Client refund/chargeback fallout, and potentially criminal and civil claims (the speaker says criminal investigation wasn’t clearly visible at the time of speaking).

---

3) The “client-disclosure” shift: contracts must disclose AI use

A major recommendation is to add AI disclaimer/disclosure clauses in both:

• Client contracts, and
• Website terms and conditions.

The speaker argues transparency is becoming legally required across jurisdictions and references:

• California’s AI transparency act (effective Jan 2026)
• New York’s AI-related requirements
• The EU AI Act disclosure expectations
• U.S. Copyright Office guidance suggesting disclosure in some copyright registration contexts

Practical point:

• Even if business owners worry clients will view AI use as “cutting corners,” the speaker frames silence as riskier than disclosure.

What disclosure should cover:

• How AI is used (e.g., drafting, editing/retouching, ideation, research, mood boards, etc.)
• While still asserting the “majority” of output is human-created.

---

4) The “two-sided” contract: protect your work from clients feeding it into AI

The speaker warns that many creative contracts don’t address what happens after deliverables are handed over.

Concern:

• Clients may upload delivered work into AI tools to generate edits, “feedback,” summaries, or new variants.

How this becomes a problem:

• Copyright law: unauthorized use of copyrighted deliverables.
• Contract breach: if the contract doesn’t prohibit such conduct.

Recommended stronger language (examples):

• Explicitly bar clients from:
  • submitting delivered work into AI tools,
  • editing/altering deliverables through AI,
  • uploading templates/course materials for summaries or regeneration,
  • and (in the strongest versions) using the client’s work for AI training or generation “in perpetuity.”

---

5) Data/confidentiality risk: don’t paste client assets into free AI tools

Another warning is about confidentiality and data protection.

Risk described:

• Uploading client brand files, unreleased materials, addresses/phone numbers, proprietary information, etc. into consumer/free AI tools may breach confidentiality obligations.
• The speaker claims these tools may use prompts/uploads in ways that could train models—or at minimum create a loss of control.

Suggested safeguards:

• Use enterprise/business-tier tools
• Enable “no training” options where available
• Get written client permission
• Document everything

---

6) AI chats are not private: prompts can be discoverable and admissible

A central message is that AI chat logs are not confidential like lawyer-client privilege.

Key points cited:

• Courts treat AI conversations as electronic records that can be discoverable and subpoenaed.
• The speaker alleges a Jan 2026 federal judge order required OpenAI to produce ~20 million ChatGPT conversation logs in a copyright lawsuit, and users were not notified.
• The speaker also references criminal and civil examples where AI chat content was used to support intent/timeline or appeared in judicial written opinions.
• Deleting a chat does not remove it from system logging; metadata persists.

---

7) Broader legal context: courts are treating AI conduct as traceable, consequential

Briefly, the speaker discusses a high-profile federal appellate ruling involving AI-generated child abuse imagery.

• The speaker says purely private possession of fully AI-generated images (not depicting real children) was treated as protected under First Amendment protections (narrowly).
• They emphasize this does not legalize production/sharing, and does not permit any involvement with real children.
• Defendants may still face serious federal charges in cases involving real children.

Purpose of this discussion:

• To show legal systems are addressing AI issues at serious levels, not as hypothetical risk.

---

8) Action steps proposed at the end

The speaker provides a “doable” checklist:

1. Add an AI disclosure/disclaimer clause to client contracts if you use AI in your work.
2. Add contract language prohibiting clients from feeding your delivered work into AI tools (and altering deliverables via AI).
3. Stop putting confidential client materials into free consumer AI tools; use protected/data-safe methods and get permission.
4. Audit your portfolio: anything AI-involved should be disclosed, and the majority should be human-created.
5. Treat AI chat tools as potentially legal evidence—assume it could be pulled into legal proceedings.

Additional recommendation:

• Buy pre-written contract language rather than having AI generate contract terms, because updates may be needed as laws change.

---

Presenters or contributors

• Paige (speaker/lawyer) — introduced repeatedly as “Paige,” associated with The Legal Page, and presented as providing legal guidance.
• “Griffith Law” / Griffasawwisaw.com — referenced as the law firm for one-on-one legal support (speaker’s firm).
• The Legal Page / legal team at The Legal Page — mentioned as assisting in Q&A and contract updates.
• Court systems / FTC / state attorneys general — referenced as external entities (not individual contributors).