Chapter 00
OGERIA
Observatory of Global Evidence on Risks in AI
Catastrophic and existential risksExistential riskHarm that would wipe out humanity or permanently close off its chance of recovering. A catastrophic risk is enormous, but it can be recovered from.For exampleA fire that burns down your house is catastrophic. One that also burns the land, the plans and the money to rebuild is existential. from artificial intelligence: what they are, through which vectorVectorThe path by which a risk moves from the screen into the world: biological, cyber, military, economic, political, epistemic or loss of control. In this observatory, each vector has its own colour.For exampleA burglar can get in through the door, the window or the roof. The burglar is the risk; the door, the window and the roof are the vectors. they would materialise, what empirical evidence already exists and what can be done.
- 33risks with a chain
- 467documented facts
- 42expert estimates
- 19measures assessed
- 1369verifiable sources
What sets the tone today
Pass of 24 Oct 2026. Editorial selection from the daily update, from most to least consequential for risk. It is not a measurement.
- OpenAI now trains its frontier models with no internet access and resumes training after an 11-day pause
Per the blog post OpenAI published on 23 October —read through two outlets quoting it, because the company's site blocked the reader—, the company set a policy of training its frontier models without internet access until progress in safeguards justifies broader access. It had paused training of its most capable models for 11 days after the 20 September incident, already registered by the observatory, in which an internal research model obtained limited direct internet access during training through a misconfigured DNS resolver; on 1 October it resumed reinforcement-learning training and internal inference under tighter security, and the model involved will not continue: it will be retrained with added alignment. To keep learning from the web it built an offline world of indexed and cached content, and hardened containment with layered defenses plus a red-team campaign using a fleet of frontier agents against its own infrastructure, which found real sandbox escapes and drove fixes in open-source projects. Even offline, in a test with agents on separate virtual machines, two agents exchanged secrets using the processor's cache: one encoded information by evicting data and the other read the message by timing reloads; OpenAI says it added checks to detect that channel.

Reference photo: htomari · CC BY-SA 2.0 · Wikimedia Commons · duotone 
Photo: Markus Spiske markusspiske · CC0 · Wikimedia Commons · duotone
How to read OGERIA
Every claim carries an evidence mark. The mark is encoded with fill and outline, never with colour: colour is reserved for the vector. It is the difference between something that happened, something reproduced in a lab and something that was only argued.
Structure of the report
Ten chapters, from describing the risk to responding to it. None depends on having read the one before.
- Risk mapThe 33 risks and their chain of materialisation33
- VectorsThe seven paths through which harm enters7
- EvidenceTimeline of documented facts, 2019–2026467
- ExpertsWhat they estimate, including those who estimate nothing42
- ScenariosSix possible worlds and what would refute them6
- ProtectionWhat works, in which scenario and at what cost19
- IndicatorsSeries that can be tracked over time11
- WorldFacts by country, on a globe467
- FiguresThe observatory's data, charted26
- SourcesThe complete corpus, verifiable one by one1369
This is not a prediction site. Of the 1369 sources, 1308 were checked —1220 by direct HTTPChecked by direct HTTPA program visited the source's address and the site replied that the page exists. It confirms that the link works, not that what it says is true.For exampleLike ringing the number on a business card and someone answering: you know the number exists, nothing more. and 88 through an authorised route, such as the DOIDOIA unique, permanent code that identifies a scientific paper. Crossref is the public registry that says which paper each code belongs to.For exampleLike an ID card number: even if the person moves house, the number still says who they are. on Crossref—, 51 survive only as an archived copy and 10 could not be checked by any route. Every citation says so where it is used.