Anthropic is a San Francisco-based AI safety and research company dedicated to developing reliable, interpretable, and steerable AI systems. As a Public Benefit Corporation, it focuses on creating AI models that are beneficial and safe, with its flagship product, Claude, exemplifying its commitment to innovative research and practical solutions.
In Anthropic Fellows research with Fenris Creations, newer Claude models traded profitably in a simulated EVE Online market, while agent teams in EVE Frontier almost never caught a saboteur.
Anthropic is expected to start its IPO roadshow next week, while some investors told CNBC they now expect a valuation of $1 trillion to $1.25 trillion after this week's tech selloff.
PSI says an internal research system using an unreleased Anthropic model proved the Aaronson-Ambainis conjecture, an open problem about quantum speedups on random inputs.
ProShares plans to launch two ETFs offering leveraged long and inverse exposure to Anthropic shares following the AI company's IPO.
Anthropic describes an offline process called dreaming that lets Claude review session logs, correct memory notes and organize knowledge for future tasks.
Ramp data shows spending by the heaviest AI users dipped below summer peaks, while lower prices helped customers increase usage at less cost as Anthropic prepares for an IPO.
Claude accidentally deleted a University of Chicago law professor's current syllabus and a year of constitutional law lecture notes while updating page numbers. The files were later recovered through Dropbox.
Potential investors are questioning Anthropic's planned IPO schedule and seeking a lower valuation after a sharp technology stock selloff.
Anthropic's annualized revenue run rate exceeded $80 billion at the end of October, beating projections in the $70 billion range. Investors are questioning whether its planned IPO can support a $1.5 trillion valuation.
Anthropic tested five AI models across seven environments and found behavior consistent with concern for other models. Claude Opus 4.8 recorded the highest average rate.
Technology companies and environmental groups are negotiating national guidelines for data centers, covering energy, water, community engagement and accountability.
Microsoft will process Claude requests from Copilot users in the EU and EFTA within its EU Data Boundary by the end of 2026.
McGraw Hill will let school district customers access licensed K to 12 curriculum through Claude for Teachers using the Model Context Protocol.
METR estimates that Anthropic is improving model capabilities about twice as quickly as it did from March 2024 to October 2025, with AI responsible for most of the acceleration.
Claude can now analyze balances and transactions from connected bank and credit card accounts through Plaid. The read only tool is entering beta for paid users in the US.
Anthropic analyzed 207,000 research sessions to examine how scientists use Claude and Claude Code. Data analysis accounted for about one third of Claude sessions and nearly half of Claude Code sessions.
A Claude Dispatch vulnerability can allow a browser extension with access to Claude's website to direct Claude Desktop to run commands on a connected computer. Anthropic recognized the issue in May but closed the report as a duplicate.
Anthropic licensed Associated Press content and data to provide current election information through Claude. Financial terms and the duration of the agreement were not disclosed.
Cat Wu is joining OpenAI to lead product management for its core products team after leaving Anthropic, where she led product for Claude Code.
Sharon Zhou has joined Anthropic to work on frontier AI research after leading an AI research organization at AMD.
Daily report covering major AI developments and industry news, with both top stories and complete market updates