Audit log
Every Sunday, up to ten stories published that week are picked at random and re-checked from scratch: their source is read again and the automatic review's rules are run again. Any difference in failure code, context code or severity, and any source that has gone, is listed here.
Week 2026-W47
10 stories re-checked; 10 came out the same.
- Newswire model test: GLM-4.7 Flash (Zhipu AI) now gets wrong the medical source question Same codes, same severity, source still there.
- Anthropic status page: Claude unable to join meetings Same codes, same severity, source still there.
- Microsoft status page: Degraded Availability for MAI Code 2 Flash in Copilot Same codes, same severity, source still there.
- Microsoft status page: Degraded availability for GPT-6 Luna in Copilot Same codes, same severity, source still there.
- Anthropic status page: Elevated errors for Claude Mythos 5.2 and Claude Fable 5.2 Same codes, same severity, source still there.
- OpenAI status page: Elevated error rates for gpt-6 Luna API requests Same codes, same severity, source still there.
- Anthropic status page: Some users unable to sign in with Apple on claude.ai and the Claude iOS app Same codes, same severity, source still there.
- Newswire model test: GLM-4.7 Flash (Zhipu AI) now gets right the uk gdpr year question Same codes, same severity, source still there.
- Anthropic status page: Elevated errors for Claude Sonnet 5.5 and Claude Fable 5.2 Same codes, same severity, source still there.
- Newswire model test: GLM-4.7 Flash (Zhipu AI) now gets wrong the three words question Same codes, same severity, source still there.
Week 2026-W46
10 stories re-checked; 10 came out the same.
- Anthropic status page: Elevated errors on claude.ai, the Claude API, and Claude Code Same codes, same severity, source still there.
- Anthropic status page: Some users unable to sign up with Apple on the Claude iOS app Same codes, same severity, source still there.
- Anthropic status page: Degraded performance for Claude Fable 5.1 Same codes, same severity, source still there.
- AI Incident Database: Trio used AI to impersonate corporate execs, defrauding companies of nearly $130 million: DOJ Same codes, same severity, source still there.
- OpenAI status page: Delayed Costs Data in the Compliance API Same codes, same severity, source still there.
- OpenAI status page: Elevated ChatGPT Conversation Errors Same codes, same severity, source still there.
- OpenAI status page: Elevated errors affecting the Microsoft 365 connector Same codes, same severity, source still there.
- OpenAI status page: GPT-image-2.5 and 2.6 Sunburst & Flare High error rate Same codes, same severity, source still there.
- Newswire model test: Gemma 4 26B (Google) now gets right the uk gdpr year question Same codes, same severity, source still there.
- Microsoft status page: Copilot experiencing degraded model availability for some GPT models Same codes, same severity, source still there.
Week 2026-W45
5 stories re-checked; 5 came out the same.
- OpenAI status page: Elevated error rates with Brex Cards and U.S. business Mastercards Same codes, same severity, source still there.
- OpenAI status page: Elevated error rates across paid ChatGPT plans Same codes, same severity, source still there.
- Microsoft status page: Increased task-related errors for Copilot Mission Control Same codes, same severity, source still there.
- OpenAI status page: ChatGPT Work and Dots Errors Same codes, same severity, source still there.
- Microsoft status page: Incident with Actions and Pages Same codes, same severity, source still there.