Society and technologysince 26 May1,100 entries

AI: capabilities, regulation, labour

Frontier-model safety failures now directly block product releases and strain regulatory guardrails, as governments move from information requests to mandatory rules while chip export controls and supply-chain pacts harden an intensifying geopolitical race.

AI text how it is made

Photo: Brett Sayles (Pexels)

This week19 entries

State of playtoday 12:03

Containment and Litigation

Anthropic's IPO prospectus says training of its next flagship model was paused for over a month over security concerns. OpenAI halted training from 20 September to 1 October after a DNS error gave an internal model internet access. California's attorney general and Arizona are pursuing OpenAI over rogue-agent incidents.

Europe and Markets

The AI Office has asked OpenAI and Anthropic how they test models for cyber risks, and Henna Virkkunen said on 16 November that formal investigations could follow. Antonio Costa proposed on 21 November a digital-revenue levy within the 2028–2034 budget. The STOXX 600 lost 2% on the week, its steepest weekly drop since March 2025.

Chips and Export Controls

A US interim rule requires licences for advanced chip transfers and for data centres serving China-headquartered firms, pressuring Southeast Asian hosts. Washington is weighing disclosure rules for US firms using Chinese AI models. Bessent's 10-day deadline for Nvidia board answers expires around 25 November.

What comes next

  • in 3 daysBessent's 10-day deadline for Nvidia board answers on export compliance

Open questions

  • Will the AI Office's information requests to OpenAI and Anthropic turn into formal AI Act enforcement cases?
  • Will the FTC demands lead to enforcement action against OpenAI or Anthropic?

Who is who8

Anthropic Frontier AI lab

Raised $81.3bn in its IPO at a valuation of about $1.5 trillion, saw shares jump 17% on its Nasdaq debut as ANT, and still faces a nonprofit lawsuit seeking to halt frontier-model development.

20 Nov

OpenAI Frontier AI lab

Published a framework on 18 November after a DNS error gave a research model internet access, introduced a Control Room it says can stop output within 35 seconds, and now faces California and Arizona actions.

20 Nov

European Commission EU executive and AI Act enforcer

Its AI Office is collecting information from OpenAI and Anthropic, no infringement finding exists, and it says an AI-assisted cyberattack hit its own systems from 22 October for at least 10 days.

20 Nov

Nvidia AI chip maker

Reported quarterly revenue of $115.8bn, created a board committee on national security and export compliance, says its China licences carried strict conditions and it has stopped H200 shipments, and faces a Treasury demand for answers.

20 Nov

Promises12

Germany, Sweden, the Netherlands and the European Commissionopen

Germany, Sweden, the Netherlands and the European Commission will join the San Francisco Declaration the week after 16 November.

16 Nov · due 23 Nov

Scott Bessentopen

Nvidia's board and advisers must answer Treasury's China export-compliance questions within 10 days of 15 November.

15 Nov · due 25 Nov

Australian governmentopen

Australia will introduce legislation next year creating an AI safety regulator with mandatory reporting of misaligned behaviour and cyber incidents.

10 Nov · during 2027 · pollar.news

Google, Microsoft, OpenAI and Anthropicopen

Google, Microsoft, OpenAI and Anthropic will honour a single robots.txt rule for opting out of generative AI training, with most changes taking effect in 2027.

31 Oct · most changes in 2027 · pollar.news

Chronicle1,100 entries since 26 May

16–22 November19 entries

21/11

Countries meeting in Geneva on 21 November agreed to continue discussions on lethal autonomous weapons systems for another two years, to 2028, without committing to negotiate firm limits or a prohibition. The proposed outcome points to an instrument under international humanitarian law.

apnews.com · politico.eu

20/11
20/11
19/11
19/11
19/11
19/11
19/11
19/11
18/11
18/11
18/11
17/11
17/11
16/11
16/11
16/11
16/11
16/11

9–15 November21 entries

15/11
15/11
15/11
13/11
13/11
13/11
12/11
12/11
12/11
12/11
12/11
11/11
11/11
11/11
10/11
10/11
10/11
10/11
10/11
09/11
09/11

AI writes this thread from the sources cited with every entry. How it works

View history

Get Pollar Week

The week in news, every Friday. Free.

Free. Unsubscribe anytime.