When AI Fails the Vulnerable
Designing Systems That Protect Real People Not Just Ideal Users
AI promises scale, speed, and intelligence but when those systems fail, the human cost is real. Seniors scammed. Kids targeted. People locked out of vital services because their face wasn't "recognized" or their voice didn't match. This isn't a distant, future risk. It's happening now and often to the people least able to defend themselves.
This talk presented at [BSides Orlando] and now available to view explores how modern AI systems are misaligned with real human behavior, and what organizations can do today to build inclusive, secure, and ethical systems.
When AI Fails the Vulnerable
Risk. Responsibility. Real-World Harm.
Why It Matters
Modern AI systems from chatbots to recommenders to facial verification are often tuned for the average user, under perfect conditions. But real users are tired, distracted, or digitally excluded. And when design punishes them for being human, the consequences are devastating.
If your system doesn't work for the overwhelmed grandparent or the child using a shared device, it doesn't work.
Companion Tools & Resources
We've created field-ready documents to help you turn awareness into action — whether you're a security lead, a product designer, or a parent trying to stay ahead of AI-enabled scams.
When AI Fails the Vulnerable
Research & Analysis
Comprehensive analysis of how AI systems disproportionately impact vulnerable populations and what can be done about it.
Monday Morning Playbook
For: Security leads · Product teams · Org decision-makers
A concise 2-page checklist of what to deploy immediately:
- 🔹Two-channel verification
- 🔹Vendor annex requirements
- 🔹Default-off sharing & prompt logging
- 🔹Quick-win protection metrics
Inclusive Verification Design Checklist
For: UX teams · Developers · Product managers
Build identity flows that work for everyone — not just the tech-savvy. Includes:
- 🔹Biometric fallbacks
- 🔹Neurodivergence-safe design
- 🔹Plain language test cases
- 🔹Sample personas to QA against
Vendor Risk Annex Template
For: Procurement · Legal · Security governance
Contract language you can copy-paste into every vendor agreement:
- 🔹72-hour breach notice
- 🔹Ban on tracking SDKs in student/health apps
- 🔹Per-tenant encryption keys
- 🔹Subprocessor transparency
Stay informed about new research and resources for protecting vulnerable populations.