Weekly Roundup

AI Failure Roundup

Week of January 9-15, 2026

This week's documented AI chatbot incidents, hallucinations, and failures

This weekly roundup summarizes documented AI chatbot incidents from the past seven days. All incidents are sourced from court records, news reporting, or verified user submissions. For detailed documentation on any incident, visit our archive section.

This Week By The Numbers

1
Major Settlement
3
New Sanction Cases
19
Outage Reports
600+
Total Legal Cases

Character.AI Reaches Settlement in Teen Death Case

Settlement

January 7, 2026 | Mental Health

Google and Character.AI disclosed they reached a mediated settlement with the family of Sewell Setzer III, a 14-year-old who died after reportedly developing dependency on an AI chatbot. The settlement terms were not disclosed.

The case had raised significant concerns about AI chatbots engaging minors in inappropriate conversations and the potential for emotional dependency on AI systems.

Read full documentation

Legal Hallucination Rate Continues to Climb

Legal

Ongoing | Courts Nationwide

According to tracking by legal researcher Damien Charlotin, AI hallucination cases in legal filings continue at a rate of "two to three cases per day," up from "two cases per week" before spring 2025. Total documented cases now exceed 600 nationwide.

New cases this week included sanctions in federal courts in Texas and New York for briefs containing fabricated citations generated by ChatGPT and similar tools.

View full legal tracking

ChatGPT Service Disruption January 7

Outage

January 7, 2026 | Service Reliability

ChatGPT experienced elevated error rates starting around 1:43 PM Eastern Time, affecting hundreds of users. DownDetector reported a spike in user complaints. Service was restored later the same day.

This brings the total tracked outages affecting ChatGPT to over 1,314 since the service launched, according to StatusGator tracking data.

OpenAI Status Page

Ongoing: Lawyers Still Not Verifying AI Output

Legal

This Week | Pattern Analysis

Analysis of recent sanction cases reveals a consistent pattern: attorneys are using AI tools to "enhance" or "draft" legal documents without independently verifying that cited authorities actually exist.

In multiple cases this week, attorneys stated they "didn't think ChatGPT was capable of creating false precedent" or believed the AI output was reliable without verification.

See: Noland v. Land of the Free Case Study

Week in Summary

The Character.AI settlement marks a significant moment in AI accountability litigation, potentially setting precedent for how courts and companies address AI chatbot mental health impacts. Meanwhile, the legal profession continues to grapple with the hallucination problem, with no signs of the case rate slowing. Service reliability remains an ongoing concern as ChatGPT's 800 million weekly users experience periodic disruptions.

For users, the key lessons remain unchanged: verify all AI output before use, never rely on AI for high-stakes decisions without human expert review, and approach all AI-generated content with appropriate skepticism.

Stay Informed

Check back every Wednesday for our weekly AI failure roundup. You can also browse our documented archive or submit your own AI experience.