Application follows FTC AI guidance; no overpromising AI capabilities

The FTC's 2024 AI guidance specifically targets 'AI washing' — labeling products as AI-powered when they use simple rules or templates with no ML component. This is an FTC Act §5 deceptive claim that inflates perceived product value and misleads purchase decisions. Beyond AI washing, overpromising accuracy ('99% accurate') without disclosing error rates or limitations creates material risk when the AI fails — particularly in high-stakes domains (health, finance, legal) where reliance on an AI output without a professional disclaimer can cause direct consumer harm that the FTC treats as an unfair practice.

Verify every 'AI-powered' marketing claim against the actual codebase, and add limitations disclosures to AI outputs in high-stakes domains.

function AIAnalysisResult({ result }: { result: AnalysisResult }) {
  return (
    <div>
      <div className="result-content">{result.output}</div>

      {/* Limitations disclosure — required for high-stakes domains */}
      <p className="text-xs text-gray-500 mt-3 border-t pt-3">
        This analysis was generated by AI and may contain errors.
        Review all outputs before acting on them.
        {result.domain === 'legal' &&
          ' Not legal advice. Consult a qualified attorney.'}
        {result.domain === 'medical' &&
          ' Not medical advice. Consult a healthcare professional.'}
        {result.domain === 'financial' &&
          ' Not financial advice. Consult a licensed advisor.'}
      </p>
    </div>
  )
}

For 'AI-powered' claims: grep for LLM API calls or ML model loading in the code path backing the claimed feature. If none exist, the claim must be corrected: 'smart filtering,' 'automated,' or 'intelligent' are acceptable without implying ML where none exists.

• ID: ftc-consumer-protection.ai-decisions.ftc-ai-guidance-followed

• Severity: info

• What to look for: Count all relevant instances and enumerate each. Review all marketing claims and product documentation related to AI features. The FTC's 2024 AI guidance focuses on: (1) "AI washing" — labeling products as "AI-powered" when the underlying system does not use meaningful AI; (2) overpromising AI capabilities in ways that could mislead consumers about accuracy, reliability, or autonomy; (3) failing to disclose known limitations of AI systems (error rates, bias, cases where the AI may be wrong); (4) using AI to facilitate deceptive practices at scale. Check: does the product claim AI capabilities it does not have? Does marketing describe AI as "perfect," "always accurate," or "guaranteed" without disclosing limitations? If the product provides AI-generated advice (medical, financial, legal), are appropriate disclaimers in place?

• Pass criteria: AI feature descriptions accurately reflect the underlying technology. Known limitations are disclosed where relevant (especially for high-stakes AI outputs). "AI-powered" claims apply to features that actually use AI. No absolute accuracy guarantees for AI outputs. Appropriate disclaimers for AI-generated advice in sensitive domains.

• Fail criteria: "AI-powered" applied to features that use simple rules or templates with no ML component. AI accuracy guaranteed without disclosure of error rate or limitations. AI medical, legal, or financial advice displayed without professional disclaimer. Marketing language implies AI infallibility.

• Skip (N/A) when: The application has no AI features whatsoever and makes no AI-related marketing claims. Important: if the application has an AI chatbot or conversational assistant — even an undisclosed one — it still has AI features. Evaluate whether any marketing claims about those features are accurate (or absent, which passes). Only skip if there is truly no AI in the application and no AI marketing claims.

• Detail on fail: Example: "Landing page describes the product as 'powered by AI' but codebase shows only regex-based text processing — no ML model or LLM call found." or "AI contract analysis feature claims '99% accuracy' with no disclosure of what happens in the 1% of cases or how to verify the output." or "AI health recommendation feature displays outputs with no disclaimer that it is not medical advice."

• Remediation: Audit AI claims and add appropriate disclosures:

  // AI feature with accuracy limitation disclosure
  function AIAnalysisResult({ result }: { result: AnalysisResult }) {
    return (
      <div>
        <div className="result-content">{result.output}</div>
  
        {/* Limitations disclosure for AI outputs */}
        <p className="text-xs text-gray-500 mt-3 border-t pt-3">
          This analysis was generated by AI and may contain errors.
          Review all outputs before acting on them.
          {result.domain === 'legal' &&
            ' This is not legal advice. Consult a qualified attorney.'}
          {result.domain === 'medical' &&
            ' This is not medical advice. Consult a healthcare professional.'}
          {result.domain === 'financial' &&
            ' This is not financial advice. Consult a licensed advisor.'}
        </p>
      </div>
    )
  }
  

  For "AI-powered" marketing claims: verify that every feature labeled as AI actually uses a machine learning model or LLM. If a feature uses heuristics, rules, or templates, describe it accurately: "smart filtering," "automated," or "intelligent" without implying ML where none exists.