Tennessee Woman Jailed for Six Months Due to AI Facial Recognition Error
Case Details
Angela Lipps, a 50-year-old mother of three and grandmother of five from north-central Tennessee, was arrested by US marshals at her home in July while babysitting four children. She was taken away at gunpoint and booked as a fugitive from justice from North Dakota.
How the AI Error Occurred
Fargo police detectives investigating bank fraud cases in April and May 2025 reviewed surveillance video showing a woman using a fake US army military ID to withdraw tens of thousands of dollars. The officers used facial recognition software to identify the suspect as Lipps. A detective wrote in court documents that Lipps appeared to match the suspect based on facial features, body type, and hairstyle.
Lipps told reporters she had never been to North Dakota and didn't know anyone there. No one from the Fargo police department contacted her before the arrest.
Legal Timeline
- Lipps remained in a Tennessee jail for nearly four months without bail while awaiting extradition
- She was charged with four counts of unauthorized use of personal identifying information and four counts of theft
- North Dakota authorities didn't transport her from Tennessee until the end of October, 108 days after her arrest
- She appeared in a North Dakota courtroom the next day
- She was released on Christmas Eve after her attorney obtained bank records showing she was in Tennessee at the time of the Fargo crimes
Consequences and Aftermath
While jailed and unable to pay bills, Lipps lost her home, her car, and her dog. Fargo police didn't pay for her trip home after release, leaving her stranded in North Dakota. Local defense attorneys covered a hotel room and food on Christmas Eve and Christmas Day, and the F5 Project non-profit helped her return to Tennessee.
Her attorney, Jay Greenwood, told reporters: "If the only thing you have is facial recognition, I might want to dig a little deeper."
Context: Other AI Errors
This isn't the first case of AI error flagging the wrong suspect. In October, an AI system apparently mistook a Baltimore high school student's bag of Doritos for a firearm and called police. Earlier this year, UK police arrested a man for a burglary in a city he'd never visited after face-scanning software confused him with another person of south Asian heritage.
📖 Read the full source: HN AI Agents
👀 See Also
Anthropic Reports Evidence of Mass Claude Distillation by AI Competitors
Anthropic has shared evidence that DeepSeek, Moonshot, and MiniMax were using approximately 24,000 fake accounts to conduct mass distillation of Claude, with over 16 million exchanges recorded.
Exploring Step 3.5 Flash: Open-Source Model for Fast Deep Reasoning
Step 3.5 Flash is an open-source foundation model designed for fast and efficient deep reasoning, utilizing a sparse Mixture of Experts architecture.
Talkie: A 13B LLM Trained Exclusively on Pre-1931 Text, Using Claude as a Judge in RL Training
Researchers released Talkie, a 13B LLM trained only on text published before 1931 (no internet, no WWII data). Claude Sonnet 4.6 was used as the judge in its online DPO reinforcement learning pipeline, and Claude Opus 4.4 generated synthetic multi-turn conversations for fine-tuning. The model can write Python code from a few in-context examples despite zero modern code in training.
AI Vendor Lock-In Escalates: Switching Models Now Costs More Than Most Expected
A Zapier survey of 542 US executives shows 90% thought they could switch AI vendors in under 4 weeks, but 58% of actual migrations failed or took far longer. Meanwhile, OpenAI raised GPT-5.2 input token pricing from $1.25 to $5.75, and Anthropic moved Claude enterprise to dynamic pricing, potentially doubling or tripling costs for heavy users.