Eidetic

AI Caveat

AI
Trust
Block

Trust and transparency component for displaying AI limitations, disclaimers, accuracy warnings, and safety notices. Essential for responsible AI interfaces.

Preview

Full caveat card with expandable details

Accuracy Notice

AI
Warning

This AI response is based on training data with a knowledge cutoff. Information may not reflect the most recent developments.

AI-generated content may contain errors

Severity Levels

Different severity levels for various situations

Disclaimer

AI
Info

This is general information and should not be considered professional advice.

Accuracy Notice

AI
Warning

AI-generated content may contain inaccuracies. Please verify important information.

Potential Bias

AI
Caution

AI responses may reflect biases present in training data.

Safety Warning

AI
Critical

Do not rely on AI for medical, legal, or safety-critical decisions.

Caveat Types

Pre-configured types for common scenarios

Limitations

AI
Info

This AI cannot access real-time data, browse the internet, or remember previous conversations.

Information May Be Outdated

AI
Warning

Information may be outdated. The AI's knowledge has a cutoff date.

As of 1/1/2025

Experimental Feature

AI
Info

This feature is experimental and may change or be discontinued.

Privacy Notice

AI
Warning

Your conversations may be used to improve the AI. Do not share sensitive personal information.

Banner Variant

Full-width notification for page-level warnings

Accuracy Notice: AI responses are generated and may contain errors. Always verify important information.

AI

Inline Variant

Embedded warning within content

Accuracy Notice

AI

The following code example is AI-generated and should be reviewed before use.

  • Check for security vulnerabilities
  • Test in a safe environment first

Compact Variant

Minimal single-line warning

AI-generated content - verify before use

Tooltip Variant

Minimal inline text

AI-generated

Without AI Indicator

Hide the AI badge

Disclaimer

Info

This is a general disclaimer that applies to all content on this platform.

Props

AICaveat component API reference

PropTypeDefaultDescription
type'accuracy' | 'limitations' | 'bias' | 'privacy' | 'outdated' | 'experimental' | 'disclaimer' | 'safety' | 'custom''disclaimer'Pre-configured caveat type
severity'info' | 'warning' | 'caution' | 'critical'Severity level (auto-set by type if not specified)
titlestringCustom title (defaults based on type)
message*stringMain caveat message
detailsstring[]Expandable list of details
linksCaveatLink[]Related links ({ label, href })
dismissiblebooleanfalseAllow dismissing the caveat
onDismiss() => voidCallback when dismissed
onLearnMore() => voidCallback for "Learn More" button
variant'banner' | 'card' | 'inline' | 'tooltip' | 'compact''card'Display variant
showAIIndicatorbooleantrueShow the AI badge
iconReact.ReactNodeCustom icon
timestampDateWhen this caveat applies (for outdated type)

Usage

Import and implementation example

import { AICaveat } from '@/blocks/ai-transparency/ai-caveat'
export default function Example() {
return (
<>
{/* Accuracy warning with details */}
<AICaveat
type="accuracy"
message="This AI response may contain inaccuracies."
details={[
'Knowledge cutoff: January 2025',
'Verify information from authoritative sources',
]}
onLearnMore={() => openHelpModal()}
dismissible
/>
{/* Banner for page-level warning */}
<AICaveat
variant="banner"
type="experimental"
message="This feature is in beta. Feedback welcome!"
/>
{/* Inline warning in content */}
<AICaveat
variant="inline"
type="safety"
severity="critical"
message="Do not use AI for medical diagnosis."
/>
{/* Compact for minimal disruption */}
<AICaveat
variant="compact"
type="accuracy"
message="AI-generated - verify before use"
/>
</>
)
}

Built With

2 components

This block uses the following UI components from the design system:

Features

Built-in functionality

  • Five variants: Banner, card, inline, compact, and tooltip displays
  • Eight caveat types: Pre-configured for accuracy, limitations, bias, privacy, etc.
  • Four severity levels: Info, warning, caution, and critical
  • Expandable details: Show additional information on demand
  • External links: Link to documentation or guidelines
  • Dismissible: Users can dismiss non-critical caveats
  • AI indicator: Badge showing this relates to AI content
  • Timestamp support: Show when information may be outdated
  • Dark mode: Full dark mode support

Accessibility

Accessibility considerations

ARIA Attributes

Uses semantic heading for titleColor-coded with text descriptionsLinks open in new tab with rel="noopener"

Keyboard Navigation

KeyAction
TabNavigate to interactive elements
Enter / SpaceExpand details or dismiss

Notes

  • Severity conveyed through text, not just color
  • Dismiss button has visible focus state
  • Consider adding role="alert" for critical caveats