Risks of AI in Design Applications
“When you invent the ship, you also invent the shipwreck.” - Paul Virilio, 1999 New capabilities create new risks. While AI makes many new things possible, it also fails in predictable ways. Knowing these predictable failures lets us manage against with domain expertise and AI fluency. This is a current list of possible shipwrecks.
10 types · Version 0.4 · 2026-05
01
Shallow Solutions
Looks good, hides flaws
The output is fluent, formatted, and confident. It looks like the answer. It may not be.
02
Homogenization
Convergence on average
Generative models produce output based on the distribution of their training data. They gravitate toward the most common patterns - which means away from anything distinctive.
03
Pseudoproductivity
Spinning wheels, burning tokens
AI collapses the time between intent and result. So we think. You can produce a polished deliverable before the underlying thinking has consolidated. Or you can get almost there fast, but take forever to finish.
04
Deskilling
Use it or lose it
Tasks that were previously learning opportunities become delegation opportunities. The cognitive muscle built through practice atrophies.
05
Black Box Rationale
How did we get here?
The output exists. The reasoning doesn't. At least, no record of it does. When AI generates a solution and a practitioner accepts it, neither may be able to explain why.
06
Bias Bleed
Two-way transfer of unstated assumptions
Bias doesn't only flow from the model's training data into the output. It flows upward from the practitioner's framing into the model's responses.
07
Constraint Blindness
Misses technical and organisational obstacles
The prototype looks right in the design tool. The constraints only become visible when engineering attempts implementation, by which time expectations are already set. Also your idea is completely illegal.
08
Empathy Gap
Lacks human context
AI produces output based on statistical patterns, not lived experience. It doesn't have access to the emotional, cultural, or institutional context that determines whether a design actually works.
09
Audit Theatre
Tests well, fails silently
When AI generates the work, AI evaluates the work, and AI fixes what AI flags, the audit loop closes on itself. The appearance of rigor is real but the rigor is not. The snake eats its tail.
Source: Design With AI — Risk Typology, Version 0.4 · 2026-05. View use cases by risk →View in Grid →