Google will let schools turn off its AI tools, but blocking AI fully cuts off Search

Google plans to make it easier for schools to restrict its AI tools for students in early 2027, the New York Times reports. There is a catch: schools that block AI entirely will lose student access to Google Search. TechCrunch notes this is because AI features are now built into the search engine. Google's own education posts describe the change as evolving its approach to school controls.

OpenAI's product lead says ChatGPT now has 1.2 billion weekly users

Adam Fry, who leads ChatGPT's consumer product, told Big Technology that ChatGPT has reached 1.2 billion weekly active users. He said it added 200 million of those in the past two months. He also said ChatGPT's search volume is now about 20% of Google's.

Google launches its AI story app Dreambeans in India, its first market outside the US

Google has launched Dreambeans, an experimental AI app, in India, TechCrunch reports. It is the app's first market outside the US. Dreambeans makes personalised illustrated stories from a user's Google services, such as Gmail, Photos and Calendar. In India it supports 13 local languages, but only one language at a time. Sensor Tower estimates the app had about 55,000 US downloads since June, with downloads falling 84% in October.

Nvidia says cities from Madrid to Malaysia are building public AI services on its tools

In its own announcement, Nvidia says it and its partners are expanding AI agents and simulation tools to cities worldwide. Examples include resident-service chatbots in more than 150 deployments across Spain, camera-based traffic agents in Wrocław, Poland, and a Malaysian firm, ITMAX, planning to support over 17,000 cameras by 2027. ITMAX says simulated signal timing cut peak-hour corridor travel times by up to 26%. Nvidia says some cities can start with workstations or a small GPU system. The news comes ahead of the Smart City Expo World Congress in Barcelona, 3–5 November.

Security firm CrowdStrike buys two AI startups, Fiddler AI and Dreadnode, for about $495M

CrowdStrike has agreed to buy Fiddler AI, which helps monitor AI agents, for about $400M, the Wall Street Journal reports. It is also buying Dreadnode, an automated tool for testing AI security, for about $95M. Both deals are mostly in cash. The Journal says companies are deploying AI agents faster than some security teams can set up controls.

Salesforce plans to buy Cresta, which makes AI for call centres, reportedly for $2.7B

Salesforce announced plans on Monday to acquire Cresta, a nine-year-old maker of AI software that automates customer contact centres. The Information, citing a source, reports that Salesforce agreed to pay $2.7B. The outlet describes the deal as part of a recent buying spree by Salesforce.

Apple expects holiday sales to grow but margins to shrink as memory chip costs rise

Apple expects revenue in its holiday quarter to rise 7% to 10% from a year earlier, above analysts' estimate of 6%, the Financial Times reports. It expects gross margins of 45% to 46%, below the 47% analysts expected, and its services outlook disappointed. Chief executive John Ternus called memory costs "unprecedented" and said Apple expects them to rise into December and beyond. Apple's inventory grew to $14.54B, up 154.3% from a year earlier, as it stockpiled memory, TechCrunch says. Tom's Hardware reports memory makers are prioritising AI data centres, tightening supply for phones and laptops.

Binance says North Korean hackers stole $6.3B using fake LinkedIn job interviews

Crypto exchange Binance disclosed on Tuesday that North Korean hackers stole $6.3B through a months-long scheme, Fortune reports. Binance says the attackers used fake job interviews on LinkedIn and an advertising campaign to deliver malicious software. Separately, the Financial Times reports that blockchain analysis firms Chainalysis and Elliptic expect the hackers to take months to launder the funds.

Google bans Android TV box maker Videostrong for preinstalling a malware backdoor

Google has cut ties with and banned Videostrong, a maker of Android TV boxes, 9to5Google reports. Google confirmed that the maker deliberately preloaded a backdoor for malware onto certified devices. The backdoor was used to control a network of hijacked devices called the XCPE botnet.

GitHub will block a risky automation trigger in public projects from 2 December

GitHub says its automation service, GitHub Actions, will block pull_request_target events by default in all public repositories from 2 December 2026, unless a policy allows them. This trigger runs with access to the main project's secrets. That means untrusted code from someone else's copy of the project could steal those secrets. Enforcement was first planned for 2 November. Before then, GitHub will run two-hour test blackouts on 9 and 16 November, 16:00–18:00 UTC. Private and internal repositories are not affected.

Everything else