Claude: weekly briefing
18 Oct – 25 Oct 2026
Top stories
- Anthropic launches Claude Duet personal agent in US research preview (CNBC, 21 Oct) — Anthropic has previewed Claude Duet, a proactive agent for small businesses and professionals, available to Pro and Max subscribers via a waitlist. The company said it 'might misread a thread', and it is competing with Meta's Muse and OpenAI's dots.
- CyberCartel reportedly used Claude and Grok agents to attack Latin American banks (Tom's Hardware, 23 Oct) — Security startup Straiker reported, via The Information, that the CyberCartel group used AI agents to automate attacks on banks and payment apps. Anthropic has reportedly blocked the group's access to Claude.
- Sweden's Government Offices to block Claude and ChatGPT on staff computers (Yahoo News, 24 Oct) — According to Sveriges Radio, Sweden's Government Offices will block ChatGPT and Claude from Monday, approving only Microsoft Copilot. Official Mikael Wettercrantz cited information security concerns about use of AI services without formal contracts.
- Elvis Presley rightsholders sue Anthropic over training on 274 songs (Rolling Stone, 25 Oct) — Authentic Brands Group, which controls Presley's music, has sued Anthropic, alleging infringement of 274 songs and seeking about $41 million. The lawsuit's claims have not been tested in court.
- Experts question Anthropic report on AI-driven changes to its workforce (The Guardian, 24 Oct) — Anthropic's report says junior technical hiring fell sharply and that staff now spend about 22% of their time managing Claude agents. Outside researchers have questioned its figures and the motives behind publishing it, ahead of a possible IPO.
Trends
Across this week's stories, three threads stand out, though not all are linked. Security and trust remain prominent: a reported criminal campaign using Claude and Grok agents against Latin American financial institutions sits alongside Sweden's decision to block Claude and ChatGPT on government staff computers, continuing the risk concerns of the previous briefing. Commercial positioning also continued, with Anthropic previewing its personal agent Claude Duet against Meta and OpenAI, and Salesforce moving to bill separately for Claude use. The workforce theme has shifted from Anthropic's own report to outside scrutiny, with experts questioning its motives. The Elvis Presley copyright lawsuit stands apart as a legal matter with no direct link to the other stories.
What it means (opinion)
Opinion (geopoliticspulse.com editorial view): The week suggests that Claude's commercial growth and its exposure to risk are rising together. Governments and criminals are both testing the product's boundaries, while Anthropic pushes new agents and, according to critics, presents its own workforce data in a favourable light before a possible listing. Next week, watch whether Sweden's restrictions are followed by other public bodies, whether Anthropic or the affected firms give fuller detail on the CyberCartel misuse, and whether Duet's usage limits and early reviews answer the questions raised about reliability. Also watch for Anthropic's response to the Guardian's unanswered request for comment and any early procedural steps in the Presley case. Each development bears on how far enterprise and public buyers will trust AI agents with sensitive work.
AI-assisted summary of public reporting; the analysis is opinion. Sources are linked above.