PostCards
All Three Major Music Publishers Are Now Suing Anthropic
Thirty-five music-publishing entities, led by Sony Music Publishing and Warner Chappell, filed a 48-page complaint on August 28 in the U.S. District Court for the Northern District of California, naming Anthropic, CEO Dario Amodei, and co-founder Benjamin Mann as individual defendants. With Universal, Concord, BMG, and Round Hill Music already in litigation, all three major music companies are now suing the Claude maker.
The complaint leans on an evidentiary trail from Anthropic's earlier authors' lawsuit, which settled in July for a record $1.5 billion after Judge William Alsup found the acquisition of roughly 5 million pirated books — not their transformative use — unlawful. The publishers argue the same books contained lyrics and sheet music that never entered any license.
Sony and Warner are seeking statutory damages of up to $150,000 per work willfully infringed, plus $25,000 for each alleged stripping of copyright management information — a figure that could run into the billions. "We disagree with the publishers' claims and we intend to defend ourselves robustly in court," an Anthropic spokesperson told TechCrunch.
DeepSeek Nears a $7.4B Round at a $74B Valuation, Eyes a 2027 IPO
DeepSeek is close to completing a roughly 50-billion-yuan ($7.4 billion) round that would value the company at about 500 billion yuan before investment, according to the Wall Street Journal and South China Morning Post. Existing backers Monolith Management, Shixiang Capital, and battery giant CATL are reportedly participating, alongside new entrants CPE and Legend Capital.
The round follows DeepSeek's first outside financing in June, which valued the company above $50 billion. It comes as the lab prepares to hire banks for a Shanghai Star Market debut, with a filing possible by year-end and a public listing targeted for 2027 — a rare path to the public markets for a Chinese frontier AI developer.
OpenAI Cuts Off Cursor's Model Access After SpaceX's Takeover
Fourteen days after SpaceX completed its $60 billion acquisition of Cursor's parent Anysphere, OpenAI notified SpaceX it will terminate direct model access, giving developers the maximum notice its contract allows. "We cannot be confident that SpaceX will use our technology within our terms of service," the company wrote, pointing to past disputes involving Twitter and xAI.
Cursor CEO Michael Truell said OpenAI models serve about 5% of Cursor's traffic and that the companies are in talks to resolve the dispute. Anthropic, whose Claude models already power the bulk of Cursor's inference, says it is increasing compute to absorb the shift — an unexpected beneficiary of a rivalry rooted in Musk's long feud with Sam Altman.
Infostealer Malware Is Hijacking Claude Sessions to Drain Usage
Anthropic says it has identified six infostealer families — Vidar, LummaC2, StealC, RedLine, and Acreed on Windows, plus Atomic Stealer on macOS — quietly copying browser passwords, login cookies, and stored credentials from infected machines. Because these tools steal already-authenticated session cookies, attackers can replay a victim's Claude session and burn through paid usage without ever touching a password or two-factor code.
"If your usage limits looked like they refilled and then drained while you weren't using Claude, this was likely the cause," Anthropic told affected users. The company stressed the malware is general-purpose and unrelated to Claude itself, typically arriving through pirated downloads or malicious apps — but the account-side fixes don't remove the infection, so a freshly issued session can be stolen again.
Chatbots Beat Search Engines at Resisting Foreign Propaganda
NPR and NewsGuard researchers built 30 questions from Kremlin-, Beijing-, and Tehran-aligned false narratives that surfaced between December 2025 and July 2026, then posed neutral and leading versions of each to ChatGPT, Gemini, Copilot, Meta AI, Grok, and Claude, comparing the results against AI summaries from Google, Bing, and DuckDuckGo.
The chatbots, tested with live web access, more consistently flagged the claims as disputed or fabricated than the search engines' summary boxes did — a rare data point suggesting that conversational grounding and citation habits may currently outperform snippet-style search summarization on adversarial, fast-moving disinformation.