Episode 28 of Impractical Privacy, hosted by Sudo, dives into the severe, real-world consequences of law enforcement's increasing reliance on flawed facial recognition algorithms. The episode highlights how this technology is structurally biased—producing significantly higher false match rates for women, the elderly, and especially people of color.

Through devastating real-life examples, Sudo explains that police are bypassing fundamental investigative work due to "automation bias," choosing to treat algorithmic guesses as undeniable truth even when confronted with blatant physical evidence to the contrary. Ultimately, the host urges listeners to push back through local advocacy, legislative bans, and physical obfuscation.

📚 Chapters

• Six Months for a Lookalike Kimberlee Williams spent six months in jail because investigators blindly trusted a false facial recognition match over her actual alibi.
• The Warning Label Fallacy Police routinely ignore software warnings, treating unverified algorithmic "leads" as definitive identifications and forcing witnesses to validate false matches.
• The Human Cost and Structural Bias Structural bias in facial recognition disproportionately misidentifies minorities, leading officers to arrest innocent people despite obvious physical discrepancies.
• What Can We Actually Do? Sudo urges listeners to combat surveillance through real-world actions like demanding legislative bans, filing FOIA requests, and using physical obfuscation.

🛠️ Resources & Tools

• ACLU Facial Recognition Case Registry
• Kimberlee Williams Case
• Randal Quran Reid Settlement
• The 2019 NIST Demographic Report (NISTIR 8280)
• Ongoing NIST Face Recognition Technology Evaluation

🌐 Connect

• Website: https://impracticalprivacy.com
  • The tracker-free, telemetry-free hub for the show, now including Bitcoin and Monero support options.
• Patreon: https://impracticalprivacy.com/patreon
• X (Twitter): @The_IP_Podcast
• Mastodon: mastodon.social/@ImpracticalPrivacy
• Bluesky: impracticalprivacy.bsky.social