Claude
Anthropic large language model family, powering chat, API and enterprise AI deployments since 2023.
Claude Fable 5 and Mythos 5 went dark for every user on 12 June, after a US deemed-export order barred foreign-national access outright, a shutdown OpenAI's comparable, equally vulnerable model never faced.
Last refreshed: 27 July 2026 · Appears in 1 active topic
Why is Claude inference now running on the same edge network that delivers Netflix streaming?
Timeline for Claude
Mentioned in: Anthropic now outspends Nvidia in Washington
AI: Jobs, Power & MoneyNamed explicitly in MOW's Search-Only Contracts as a target crawler subject to the £500 per-article charge
Media's AI Pivot: Publishers bill AI £500 a pieceMentioned in: Washington pulls a live AI model
AI: Jobs, Power & MoneyMentioned in: CAIDA due before College, scope cut
European Tech SovereigntyMentioned in: Mistral wins Airbus and BMW on merit
European Tech SovereigntyBackground
Claude is the large language model family built by Anthropic, the San Francisco AI lab Dario and Daniela Amodei founded in 2021. It launched in March 2023 and has shipped successive generations through Claude 2, the Claude 3 family, Claude 3.5 Sonnet and the Claude 4 line, offered both as the Claude.ai consumer chat product and as an API for enterprise and developer deployment.
Claude competes with OpenAI's GPT family, Google's Gemini, Meta's Llama and xAI's Grok. That competitive position sits alongside Claude's own regulatory exposure: Anthropic's federal lobbying spend hit $1.97 million in the second quarter of 2026, up 26% and enough to overtake Nvidia's outlay, disclosed in the same window as the Commerce Department's suspension order against Claude itself.
Claude's unlicensed training corpus sits inside the Bartz v Anthropic class action, whose proposed $1.5 billion settlement is under federal court review and is expected to set a per-work licensing floor other AI labs will use to anchor bilateral deals with publishers.
Export order forced Claude offline
Claude Fable 5 and Mythos 5 went dark for every user worldwide on 12 June 2026, an outage rooted in policy rather than a technical fault: neither model can verify a user's nationality mid-conversation, so a US bar on foreign-national access Left no way to comply short of a blanket shutdown. Enterprise customers running Claude inside regulated workflows, banks, cybersecurity firms and public agencies among them, lost access without warning, even though the flagged jailbreak had originated in a competitor's product.
Service partially resumed on 27 June, but the interruption Left a mark distinct from the regulatory story behind it: OpenAI's comparable model, carrying the identical vulnerability, never Left the shelf, meaning Claude alone absorbed 15 days of lost usage and enterprise trust while a rival with the same flaw kept trading.
Its training data faces a $1.5bn bill
Judge Martinez-Olguin took the $1.5 billion Bartz v Anthropic copyright settlement under submission on 18 May 2026, then ordered supplemental briefing on late opt-outs due 21 May. The class claimed 447,576 of 482,460 works, and lawyers filed for $187.5 million in fees, treating AI copyright class actions as investable.
A final settlement sets a per-work licensing floor other AI labs will use to anchor bilateral deals, a cost pressure Claude will feel directly as publishers experiment with their own pricing: the UK's Movement for an Open Web began charging AI crawlers £500 per article copied without permission from 15 June, though no test case has yet established whether such terms bind.
What is Claude AI?
How is Claude used in media and journalism?
Where does Claude inference run?
