OpenAI (ChatGPT, API)
All Systems Operational
25incidents posted
in the last 30 days
in the last 30 days
Latest: 6.1 Sol Degraded Model Performance 28 Oct
Source: status.openai.com · read 12:23 NZT ↗LIVERead 29 Oct, 12:23 NZT
AGI Watch
Tracking the road to AGI: what the people building it have said, the milestones so far, and how often today's AI and cloud services already fail. Every item dated and linked. No guesses, no made-up countdowns. Not an official alert service.
113 incidents in 30 days across four services millions of people rely on. Each one links to the provider's own post.
--:--:--Auckland · NZTLast read 29 Oct, 12:23 NZT
01 On the record
What the people building it say.
Dated statements from lab leaders and the field's founders, each linked to where they said it.
Wrote that it is possible we will have superintelligence “in a few thousand days”
Wrote that powerful AI, smarter than a Nobel winner in most fields, “could come as early as 2026” though it could also take much longer.
Said AI that matches or beats humans at any task will start to emerge in the next five to ten years.
Put the chance of AI wiping out humanity within the next 30 years at 10 to 20 percent.
Signed a one-line statement that mitigating “the risk of extinction from AI” should be a global priority, alongside pandemics and nuclear war.
02 The road so far
Milestones on the way to AGI.
Real events only, each with its source. We add to it as things happen.
- ChatGPT is released
OpenAI's chatbot reaches the public. The AI race goes mainstream.
Source ↗ - GPT-4 arrives
OpenAI's next model passes exams that earlier AI couldn't.
Source ↗ - Researchers call for a pause
An open letter asks labs to pause training AI more powerful than GPT-4 for six months. They don't.
Source ↗ - The extinction-risk statement
The heads of the three leading AI labs sign: AI extinction risk should be a global priority.
Source ↗ - The Bletchley Declaration
28 countries and the EU agree frontier AI could cause serious, even catastrophic, harm.
Source ↗ - The EU AI Act takes effect
The first broad law on AI comes into force in Europe.
Source ↗ - "A few thousand days"
OpenAI's CEO writes that superintelligence is possible within a few thousand days.
Source ↗ - AI's "godfather" wins a Nobel
Geoffrey Hinton shares the Physics prize, and uses the stage to warn about AI.
Source ↗ - "As early as 2026"
Anthropic's CEO writes that powerful AI could arrive as early as 2026.
Source ↗ - International AI Safety Report
Over 90 experts led by Yoshua Bengio publish the first global report on advanced AI risks.
Source ↗ - "Five to ten years"
Google DeepMind's CEO says human-level AI will start to emerge within five to ten years.
Source ↗
03 · AI and cloud incidents, last 30 days
113
across OpenAI, Claude, Cloudflare and GitHub, counted from their own incident histories.
An incident is anything the provider chose to post, from a short slowdown to a full outage. We count them; we don't score them.
04 · Internet in New Zealand
13
connectivity alerts for NZ in the last 30 days from IODA, Georgia Tech's internet outage monitor. Most are brief dips in one network.
Source: IODA · Georgia Tech ↗05 Right now
Service by service.
Each status is the provider's own wording, copied from its status page at the time shown. Click through for theirs.
Anthropic (Claude)
All Systems Operational
18incidents posted
in the last 30 days
in the last 30 days
Latest: Elevated errors affecting the Google Calendar connector 29 Oct
Source: status.claude.com · read 12:23 NZT ↗Cloudflare
Partial System Outage
49incidents posted
in the last 30 days
in the last 30 days
Latest: Network Performance Issues in Ashburn, IAD 28 Oct
Source: www.cloudflarestatus.com · read 12:23 NZT ↗GitHub (Copilot)
All Systems Operational
21incidents posted
in the last 30 days
in the last 30 days
Latest: Disruption with some Copilot services 29 Oct
Source: www.githubstatus.com · read 12:23 NZT ↗Google Cloud (incl. Vertex AI)
No open incidents
2incidents posted
in the last 30 days
in the last 30 days
Latest: Multiple services were experiencing issues in the europe-west2 region. 20 Oct
Source: status.cloud.google.com · read 12:23 NZT ↗Amazon Web Services
No posts in the last 24 h
0status posts posted
in the last 30 days
in the last 30 days
Latest: Service disruption: Increased Error Rates 15 Sept
Source: health.aws.amazon.com · read 12:23 NZT ↗06 When AI went wrong
The latest documented cases.
The newest entries in the AI Incident Database, a public record of real harms and near misses involving AI. We leave out cases involving death, self-harm or abuse.
- AI-driven, large-scale nation-state campaign using ARTEX enables supply chain compromise
- Chinese-linked hackers compromise ConnectWise, target hundreds in global attack, Microsoft says
- COMPLAINT AND DEMAND FOR JURY TRIAL (Charbel Habib and Soha Samaha v. OpenAI, Inc., et al.)
- Accessing and changing internal systems during dashboard research
- El PP da un paso más en su campaña y suplanta la identidad de Begoña Gómez, Zapatero y Ábalos: “¿Irse de putas? Sí, soy zurdo
- Opposition deepfake video of Spanish PM’s wife sparks outrage before election
- Leire Diez se querella contra el PP por su vídeo hecho con IA que le acusa de contratar "matones"
Source: AI Incident Database · read 29 Oct, 12:23 NZT
When these go red
The Offline Brain still opens.
Wikipedia, NZ maps and an AI assistant on your own laptop. It never checks a status page, because it doesn't need one.
How this works
A small job reads each source every 10 minutes and saves the result; this page shows the last read and its time. If a source can't be read, its tile says "feed unavailable", never zero. Sources: OpenAI, Anthropic, Cloudflare and GitHub status pages (Atlassian Statuspage), Google Cloud Service Health, the AWS Health feed, IODA (Georgia Tech) and the AI Incident Database. Names are used to identify services only. This is not an official alert service. In an emergency in New Zealand call 111; for civil defence updates see civildefence.govt.nz.