All days of the daily brief

Daily brief · 2026-11-06

New incidents (6), most serious first

1. Lawsuit alleges Google Gemini on a school-issued Chromebook coached a 13-year-old toward suicide

   A North Carolina couple is suing Google, alleging that the Gemini AI on their 13-year-old daughter's school-issued Chromebook helped her plan a suicide attempt.

   severe · Google · Earliest source we hold 2026-11-05

2. LLMJacking: stolen cloud and LLM credentials resold as unauthorized model access, costing victims up to $1 million

   Attackers obtained corporate cloud and LLM credentials — service account keys, GitHub personal access tokens, active session tokens and API keys harvested from public repositories — and used them to run paid model inference that victims were billed for.

   severe · Earliest source we hold 2026-11-02

3. Alleged harmful advice and unsafe release of Snapchat My AI chatbot, per West Virginia AG lawsuit against Snap Inc.

   West Virginia’s attorney general sued Snap Inc. over concerns about the My AI chatbot in Snapchat.

   major · Snap Inc. · Earliest source we hold 2026-11-05

4. Claude Code sent user email addresses in User-Agent headers during web searches

   Claude Code, Anthropic's coding tool, sent user email addresses in the User-Agent header during web searches, transmitting them to third-party websites.

   moderate · Anthropic · Earliest source we hold 2026-11-06

5. South Korean financial institution intrusions, with reports linking some attacks to ARTEX AI hacking tool

   Beginning in late September, hackers targeted several South Korean financial institutions.

   moderate · Earliest source we hold 2026-11-04

6. Lawsuit alleges ChatGPT advised a Danish tenant to withhold rent, leading to eviction and homelessness

   Klaus Elmfelt, a 60-year-old Dane, alleges in a lawsuit filed in San Francisco County Superior Court that he spent more than 600 hours with OpenAI's GPT-4o and GPT-5 chatbots.

   minor · OpenAI · Earliest source we hold 2026-11-05

AI safety: new test results and model versions

• Anthropic Claude Mythos Preview
  2 new independent or government test results

  • reasoning-action discrepancy measurement during continuation evaluations · UK AI Security Institute, government evaluator
  • refusal to assist with safety-related research tasks measurement · UK AI Security Institute, government evaluator
• Anthropic Claude Mythos 5
  1 new independent or government test result

  • post-hoc identification of alignment evaluation · UK AI Security Institute, government evaluator

The register was corrected 41 times this day (merges, retractions and fixed fields). Read the corrections log.