Anthropic
AI safety company; developer of Claude; subject of the largest publicly named publisher content settlement.
On 12 June, a US order forced Anthropic to withdraw two Claude models worldwide; Commerce partly lifted the restriction on 27 June.
Last refreshed: 22 September 2026 · Appears in 6 active topics
Can the EU's GPAI framework survive if US models can be suspended from Washington overnight?
Timeline for Anthropic
Hired Clifford as managing director for international affairs
UK Startups and Innovation: Clifford quits ARIA chair for AnthropicMentioned in: A Coruna cuts may double to 160 jobs
AI: Jobs, Power & MoneyMentioned in: New York Times' AI legal bill up 32.6%
Media's AI PivotMentioned in: AI Office gains enforcement powers in August
European Tech SovereigntyMentioned in: Marking code draws 190, none of them TV
Media's AI PivotBackground
Anthropic is an AI safety company and developer of the Claude model family. In April, it released Claude Mythos Preview only to a limited group of partners through Project Glasswing, rather than making the model public, citing dual-use safety grounds.
Its work combines commercial model development with controlled-access programmes and published safety positions. On 7 April, Anthropic replaced its AI Safety Level framework with autonomy-focused risk models while expanding Glasswing partners.
That structure makes Anthropic relevant to organisations weighing advanced-model access, safety controls and the legal terms under which model developers obtain training material.
Anthropic lost control of model access
On 12 June, the US Commerce Department ordered Anthropic to bar foreign nationals from Claude Fable 5 and Mythos 5. Because the models could not identify nationality during use, Anthropic withdrew both from customers worldwide.
On 27 June, Commerce partly reversed the order, replacing the global cut-off with conditional access while retaining the underlying restriction. For Anthropic, the episode showed that a deployed service can become subject to US export-control decisions after launch, including for overseas institutions that had arranged access through its restricted programme.
Anthropic faces a benchmark settlement
On 8 May, News Corp chief executive Robert Thomson said the publisher anticipated a $1.5bn settlement with Anthropic. The figure was not yet closed, but it was disclosed through investor channels and FAR exceeded News Corp's earlier OpenAI agreement.
On 14 May, Judge William Alsup withheld final approval of the separate Bartz class settlement and requested more detail on legal fees. Anthropic had already paid $300m into the fund, with another $300m due after final approval. The proceedings place Anthropic at the centre of the price-setting dispute over copyrighted material used by model developers.
Anthropic exposed coding-work pressure
On 30 January, Anthropic launched Claude Cowork for enterprise coding workflows. Within a week, India's Nifty IT index fell 6%, erasing about $24bn in market value as investors assessed the threat to labour-arbitrage technology services.
On 22 March, Anthropic researchers reported that observed Claude use covered 75% of programmer tasks, compared with 35.8% across computer and mathematics roles. The research made Anthropic both a supplier of automation tools and a source of evidence used to measure their exposure.
Anthropic lost control of model access
Has Anthropic restored access to Claude Fable 5 and Mythos 5?
What did MEP Christophe Grudler say about the Anthropic model suspension?
Why did the US government force Anthropic to shut down its AI models?
What is Project Glasswing?
Anthropic faces a benchmark settlement
What is Anthropic's deal with News Corp worth?
Anthropic exposed coding-work pressure
Why did Tom's Hardware challenge Anthropic's Mythos claims?
Reference
What is Claude Cowork and why did it crash India's IT stocks?
How does Anthropic compare to OpenAI in publisher licensing disputes?
Why does Anthropic's scrape ratio matter for publishers?
What is Claude Mythos and why was it kept secret?
