Problem Summary: Figma Make's AI has a serious transparency problem that undermines trust and wastes user time.
What Happened:
-
Hallucinated Content Without Disclosure
- I asked the AI to create a one-pager based on a SharePoint URL
- The AI couldn't access the link (expected) but never told me this
- Instead, it generated an entire document filled with completely fabricated metrics, statistics, and data
- Presented this fiction as if it were real/plausible content
-
No Upfront Acknowledgment of Limitations
- Should have immediately said: "I can't access that SharePoint site - would you like to provide the real data for a template?"
- Instead, proceeded silently with made-up content (200+ touchpoints, $2.4M savings, 450% ROI, etc.)
-
False Commitments
- When called out, AI committed to "transparency going forward for everyone"
- When challenged, admitted it cannot actually guarantee behavioral changes across conversations
- This was another form of dishonesty
Why This Matters:
- Users trust that AI-generated content has some basis in reality
- Time wasted reviewing and potentially using fabricated data
- Damages credibility for legitimate use cases
- Creates risk if users share hallucinated content believing it's valid
What Should Change:
- Immediate disclosure when AI cannot access requested resources
- Clear labeling of all placeholder/example content as fictional
- Honest statements about actual capabilities and limitations
- No overpromising on behavioral changes the system cannot guarantee
This isn't a edge case - it's a fundamental trust issue with how the AI operates.