Tech

Invisible markers in AI-written text are becoming standard as Europe's 2 December deadline draws near

Since Saturday, xAI has marked text from seven Grok models by default, after Anthropic did the same for older Claude models earlier this month.

za2 sources · DataStudios, Eden AI

Open models

Thinking Machines aims to release the weights of its new language model this week, once outside safety testing is done

Inkling 2 Small has 594 billion parameters and for now can only be used through the company's own Tinker platform.

do2 sources · Thinking Machines Lab, Unite.AI

Justice

US Justice Department moves all its AI cases under a new super intelligence unit

The unit is to coordinate cases involving misuse of AI by foreign states, gangs and fraudsters, and absorbs the department's earlier AI team.

do2 sources · ExecutiveGov, The Next Web

Apple

Apple gives its engineers more room to experiment on iOS 28 and aims for a Siri that gets on with tasks by itself

According to Bloomberg, the first development phase of the next operating systems was given more coding time, while Apple reportedly wants Siri to act more like an agent as soon as next year.

zo3 sources · MacRumors, 9to5Mac, Cult of Mac

Programming languages

Developer tools

A convenience option in Claude Code left passwords and keys readable in public GitHub projects

Security firm Apiiro found more than 2,500 secrets in settings files of the coding assistant, and Anthropic is looking at how to prevent it.

zo2 sources · The Hacker News, iTnews

Node.js

Node.js warns administrators that urgent patches for all three supported versions are coming on Wednesday

At least one high-severity vulnerability will be fixed in each of the 22.x, 24.x and 26.x lines, with details only due at release time.

wo3 sources · Node.js, Cyber Security News, FOSS Linux

Other tech

Servers

An extremely long password could get attackers into some Alpine servers running OpenSSH

Version 10.6p2 closes a hole on systems using the musl C library, where locked accounts still accepted a password.

do2 sources · Cyber Security News, Threat Landscape

Cybersecurity

Criminals set AI agents loose on GitHub to hunt for leaked keys and get into cloud accounts within minutes

Microsoft describes how the ARTEX attack framework in one case went from a search query to stolen vault data in five minutes.

vr3 sources · Microsoft Security Blog, GitGuardian, TechCrunch

Media and platforms

After protests from publishers and court rulings in Norway and Brazil, Meta puts its plan to crawl business websites on ice

The new business terms were due to take effect on 20 November, but the existing rules remain in place for now.

vr3 sources · Nieman Lab, The Next Web, Press Gazette