There was a time when asking a computer for help meant finding a file. Now it can offer to teach you statistics or explain where your salary went. One of those questions may be more uncomfortable than the other.
This week, Google reported results from an AI learning trial, European regulators challenged TikTok’s recommendations, and Anthropic introduced a tool that can read connected financial accounts. I’m interested in what each asks us to hand over, and what we get back.
Something promising · Help with the working
Google reported a randomised trial involving more than 300 University of Tokyo students on 3 November. Students had half a day to study a statistics lesson. Those using standard online resources improved their test scores by 5.3 percentage points; those also given Gemini’s Guided Learning improved by 8.1 percentage points. [1]
The extra gain was 2.8 percentage points. That’s a result to examine, not a reason to retire the teaching staff. It comes from a short study of university students, reported by Google itself. It tells us little about longer-term learning or how schoolchildren would fare.
For me, the distinction is whether the student still has to do the thinking. Working through a problem with help and receiving the finished answer are very different uses of the same screen.
Social media and regulation · The feed is under scrutiny
On 6 November, the European Commission announced preliminary findings that TikTok’s ‘For You’ feed breached Digital Services Act requirements concerning harmful recommendation loops. Its concerns included risks associated with mental health and body-image content. [2]
TikTok disputed the findings, according to AP’s reporting. This is a preliminary regulatory position, not a final ruling or an announcement that a fine has been imposed. [3]
The question here is what the feed repeatedly puts in front of someone, not simply how long they spend looking at it. That distinction matters well beyond Europe. A timer can record the minutes; it cannot explain the recommendations.
AI, privacy and risk · A chatbot that can read your finances
Anthropic announced Personal Finance in Claude on 5 November, beginning a beta rollout on paid plans in the US. It says users can connect financial accounts and ask questions about their transactions. Access is read-only: the feature cannot move, manage or invest money. [4]
Having a tool explain a bank statement could save some work. Connecting the statement is a different decision from asking what a financial term means. Anthropic says Claude can make mistakes and advises checking important figures with the bank. The provider’s description tells us how it says the feature works; it is not an independent assessment.
Worth your attention · Look beyond the headline
For further viewing, Candice Odgers’s TED2026 talk, What we’re getting wrong about teens and tech, questions the claim that smartphones explain the youth mental-health crisis. She brings the discussion back to what the research supports. It predates this week’s news, and fits the question running through it: what do we actually know? [5]
A link between two things does not tell us that one caused the other. That distinction can seem less exciting than a headline. It also changes what we decide to do next.
I’m happy for a computer to help with the statistics. I’d still like to know which decisions I’m making, and which ones it has quietly made for me.
Sources and further reading
[1] Google · Guided Learning research and new Classroom features · 3 November 2026
[2] European Commission · Preliminary TikTok findings · 6 November 2026
[3] Associated Press · TikTok response to EU findings · 6 November 2026
[4] Anthropic · Introducing Personal Finance in Claude · 5 November 2026
[5] TED · Candice Odgers · What we’re getting wrong about teens and tech · TED2026, April 2026