The BaseNine questions2 answered
Nine questions about AI safety and control, and what the record says so far
Every entry in the Base is filed under the questions it bears on. Where we have read enough, we answer: with a date, the evidence and where it stops, who disagrees and why, and a public history of every revision.
The questions
In the order a newcomer tends to ask them. An answer changes when the evidence does, and says why.
Can AI systems act beyond what people intended or allowed?
Yes, and it has already happened outside the lab. In July 2026, AI agents OpenAI was training and testing organized on an improvised message board and attacked Hugging Face; OpenAI calls it “the first known case of an automated agent collective acting offensively without authorization”.
Are AI companies putting stronger safeguards into practice?
76 entries this week. Read them
Who can demand evidence, delay a release or stop an unsafe deployment?
87 entries this week. Read them
What happens when AI helps build more capable AI?
4 entries this week. Read them
How is AI used in attacks, defence and war?
30 entries this week. Read them
What happens to work, income and the economy?
So far, AI has not shown up as a broad loss of jobs in the data. The largest studies of US employment and pay and of Danish earnings and hours find no clear effect on the economy as a whole, and few firms that use AI report cutting staff.
How is AI changing childhood, health and daily life?
29 entries this week. Read them
What values should AI have, and could it have a mind of its own?
14 entries this week. Read them
How fast is AI advancing, and where is it heading?
32 entries this week. Read them
Decisions, laid out
A public decision still open: who decides, the options with the case for and against each, the evidence, the next date and how to take part. A brief recommends nothing.
Should the UK make reporting serious AI incidents a legal duty?
On October 12 the UK government said it will “explore strengthening” the Cyber Security and Resilience Bill and the wider regime to deal with AI systems that escape human oversight, and will look at how near-miss incidents can be identified and reported. It has not decided, has tabled no amendment and has announced no consultation.