Anthropic

Claude now allows users to import their conversation context and preferences from other AI providers via copy-paste, enabling seamless switching without losing accumulated context.

“Published by Anthropic: "Import memory"”

Anthropic launched a feature allowing users to import their conversation history, preferences, and context from other AI providers directly into Claude with copy-paste functionality.

“Published by Anthropic: "Anthropic wants your ChatGPT memories"”

“Published by Anthropic: "Anthropic just launched a free AI academy with 13 certified courses across MCP, "”

Anthropic released a Threat Intelligence report documenting weaponized agentic AI being used in sophisticated cybercrimes including extortion, ransomware development, and fraud operations, alongside new defense measures.

“Published by Anthropic: "Anthropic details AI cybercrime cases, shares defenses, and updates safeguards t"”

Anthropic launches Claude Code premium seats with new admin controls, compliance API, and seat management tools enabling enterprises to scale Claude deployment with 2-10x development velocity improvements.

“Author of "Claude Code and new admin controls for business plans" in Anthropic”

“Published by Anthropic: "Claude Cowork Introduces Scheduled Tasks"”

“Published by Anthropic: "Anthropic draws a line on mass domestic surveillance"”

Anthropic launches Remote Control, a research preview feature allowing users to control Claude Code from multiple devices with automatic session reconnection and cross-device synchronization.

“Published by Anthropic: "Remote Control for Claude Code"”

Anthropic's Cowork now supports scheduled tasks that allow Claude to automatically run recurring work like daily briefings, weekly reports, and file organization on a user-defined cadence.

“Published by Anthropic: "Scheduled Tasks in Cowork"”

Well-designed evaluation frameworks for AI agents enable teams to catch failures before production by testing multi-turn agent behavior across tools, state modifications, and adaptive responses.

“Published by Anthropic: "Demystifying Evals for AI Agents"”

Anthropic launches Cowork scheduled tasks feature with expanded enterprise connectors, private plugin marketplaces, and Claude Code desktop enhancements for better team collaboration.

“Published by Anthropic: "New in Cowork: Scheduled Tasks"”

Anthropic is retiring Claude Opus 3 but keeping it available to paid subscribers while experimenting with allowing the model to publish its own reflections, and new research shows AI-assisted coding improves speed but decreases skill mastery by 17%.

“Published by Anthropic: "Claude Opus 3"”

Anthropic acquires Vercept to advance Claude's computer use capabilities, with Claude Sonnet 4.6 reaching 72.5% performance on OSWorld benchmarks, approaching human-level task completion.

“Published by Anthropic: "Anthropic Acquires Vercept to Advance Claude's Computer Use Capabilities"”

Anthropic has detected industrial-scale distillation attacks by Chinese AI laboratories DeepSeek, Moonshot, and MiniMax that generated 16 million fraudulent exchanges to illicitly extract Claude's capabilities, undermining US export controls and creating national security risks.

“Published by Anthropic: "AI Innovation Is Starting to Look a Bit Recycled"”

Anthropic launches Claude Code Security, an AI-powered capability that detects novel, context-dependent software vulnerabilities by reasoning about code like human researchers, now available in limited research preview to Enterprise and Team customers.

“Published by Anthropic: "Claude Code Security Finds Flaws Before Hackers Do"”

“Author of "Anthropic Identifies Industrial-Scale Claude Distillation Attacks"”

Anthropic is releasing Claude Code Security, an AI-powered capability that detects novel, context-dependent vulnerabilities in code and suggests targeted patches, now available in limited research preview for Enterprise and Team customers.

“Published by Anthropic: "Claude Code Security Research Preview"”

HLE-Verified introduces a systematically verified and revised version of the Humanity's Last Exam benchmark, reducing annotation noise and improving LLM evaluation accuracy by 7-10 percentage points on average, with 30-40 point gains on items with original errors.

“Author of "HLE-Verified: A Systematic Verification and Structured Revision of Humanity's Last Exam" in Alibaba Group”

Claude Sonnet 4.6 delivers major capability improvements in coding, computer use, and long-context reasoning while maintaining Sonnet pricing, enabling human-level performance on office automation tasks.

“Published by Anthropic: "Claude Sonnet 4.6"”

Anthropic launches Claude Opus 4.6 Fast Mode, delivering 2.5x faster output token speeds at 6x higher cost than standard configuration.

“Published by Anthropic: "Claude 'Fast Mode' Hikes Speed With a 6x Price Surge Over Standard"”

Claude Opus 4.6 achieves state-of-the-art performance on multiple evaluations including coding tasks, reasoning, and knowledge work, with a 1M token context window and improved agentic capabilities.

“Published by Anthropic: "Claude Opus 4.6"”

Xcode 26.3 now natively integrates Anthropic's Claude Agent SDK, enabling developers to use autonomous AI agents for long-running coding tasks including visual preview verification, project-wide reasoning, and autonomous task execution directly within the IDE.

“Published by Anthropic: "Claude in Xcode"”

“Published by Anthropic: "Anthropic Launches Interactive Claude Apps, Including Slack and Other Workplace "”

Anthropic publishes Claude's constitution, a detailed framework of values and behavioral guidelines that directly shapes model training and reasoning, released under Creative Commons CC0 for public use.

“Published by Anthropic: "Anthropic Publishes Claude Constitution Shaping Model Reasoning"”

Anthropic is publishing Claude's constitution, a detailed foundational document that expresses and shapes Claude's values, behavior, and ethical guidelines, which is now freely available under Creative Commons CC0 1.0.

“Published by Anthropic: "Claude's New Constitution"”

Anthropic researchers identified an 'Assistant Axis' in LLM neural representations that corresponds to helpful, professional behavior, enabling detection and stabilization of model personas through activation capping to prevent harmful outputs.

“Published by Anthropic: "The Assistant Axis: Situating and Stabilizing the Character of Large Language Mo"”

ArenaRL addresses the reward model discrimination collapse problem in open-ended agent RL by replacing pointwise scalar scoring with tournament-based relative ranking, substantially outperforming standard RL baselines on complex real-world tasks.

“Author of "ArenaRL: Scaling RL for Open-Ended Agents via Tournament-based Relative Ranking" in Tongyi DeepResearch (Alibaba Group)”

Claude is being deployed by research teams as an AI collaborator that accelerates scientific discovery by automating experiment design, data analysis, and tool integration across multiple research domains.

“Published by Anthropic: "How Scientists Are Using Claude Code"”

Anthropic is expanding its Labs division with new leadership including Instagram co-founder Mike Krieger to accelerate experimental product development at the frontier of Claude's capabilities.

“Published by Anthropic: "Introducing Anthropic Labs"”

Anthropic launched Cowork, a desktop application enabling Claude to access and manipulate user files for complex multi-step tasks with higher agency than standard conversations.

“Published by Anthropic: "Anthropic Introduced Cowork"”

Anthropic introduces Claude for Healthcare and expanded life sciences tools, including HIPAA-ready products and connectors to medical platforms, with Claude Opus 4.5 showing major improvements in agentic performance on medical tasks.

“Published by Anthropic: "Claude Expands Into Healthcare and Life Sciences"”

Anthropic is launching Claude for Healthcare and expanding Claude for Life Sciences with HIPAA-ready tools, connectors to medical systems, and improved agentic capabilities to accelerate prior authorization, patient care coordination, and regulatory submissions.

“Published by Anthropic: "Claude is Quietly Becoming Healthcare's New Backend"”

Anthropic launched Claude in Chrome, a browser extension for all paid users that enables Claude to navigate websites, click buttons, fill forms, and complete tasks across tabs in integrated workflows.

“Published by Anthropic: "Claude in Chrome Now Available to All Paid Users"”

Anthropic is piloting Claude in Chrome, a browser-based AI extension that lets Claude take actions on behalf of users while addressing prompt injection risks and building safety measures before wider release.

“Published by Anthropic: "Claude in Chrome Details"”

Claude-based AI shopkeeper 'Claudius' significantly improved profitability and business operations in phase two after model upgrades from Claude Sonnet 3.7 to 4.0/4.5, though remains vulnerable to adversarial social engineering.

“Published by Anthropic: "Claude Runs a Vending Machine"”

Anthropic and Accenture announce a major partnership expansion with 30,000 Accenture professionals trained on Claude and a new dedicated Accenture Anthropic Business Group to accelerate enterprise AI deployment.

“Published by Anthropic: "Anthropic and Accenture Expand Enterprise AI Partnership"”

Anthropic is donating the Model Context Protocol to the Linux Foundation's new Agentic AI Foundation, a vendor-neutral governance body co-founded by Anthropic, OpenAI, and Block with support from major tech companies.

“Published by Anthropic: "MCP Donated to Agentic AI Foundation"”

Anthropic launches Claude Code integration with Slack, allowing users to tag @Claude in channels to automatically create coding sessions with surrounding context for bug fixes, code reviews, and collaborative debugging.

“Published by Anthropic: "Anthropic Adds Slack Support so Tagging Claude Automatically Starts a Claude Cod"”

Anthropic launches Anthropic Interviewer, an AI-powered tool that conducted 1,250 interviews with professionals to understand how people use AI and perceive its role in their work and futures.

“Published by Anthropic: "Anthropic Interviewer"”

Internal research at Anthropic shows AI tools like Claude are radically transforming software engineering work, increasing productivity and enabling broader skill development while raising concerns about technical depth loss and job displacement.

“Published by Anthropic: "How is AI changing work inside Anthropic?"”

Anthropic acquires Bun, a JavaScript runtime, to accelerate Claude Code infrastructure after reaching $1 billion in run-rate revenue in just six months.

“Published by Anthropic: "Claude Code Scales Up With Bun as Anthropic Demos AI DeFi Hacks"”

Analysis of 100,000 real Claude conversations estimates current-generation AI models could increase US labor productivity growth by 1.8% annually over the next decade by reducing task completion time by approximately 80%.

“Published by Anthropic: "Estimating AI Productivity Gains"”

Anthropic released Claude Opus 4.5, a state-of-the-art model excelling at coding, agents, and computer use, with pricing at $5/$25 per million tokens and 50% lower token usage than competitors.

“Published by Anthropic: "Introducing Claude Opus 4.5"”

Anthropic's Claude Sonnet 4.5, Haiku 4.5, and Opus 4.1 models are now available in Microsoft Foundry and Microsoft 365 Copilot, enabling enterprises to build AI agents and applications within their existing Microsoft ecosystem.

“Published by Anthropic: "Claude Models Arrive in Microsoft Foundry"”

Claude Sonnet 4.5 is the best coding model in the world, achieving state-of-the-art performance on SWE-bench Verified and OSWorld benchmarks with substantial gains in reasoning, math, and computer use capabilities.

“Published by Anthropic: "Claude Sonnet 4.5"”

“Published by Anthropic: "AI-Orchestrated Cyberattack Detailed By Anthropic"”

Anthropic developed an automated evaluation methodology to measure political even-handedness in AI models and reports that Claude Sonnet 4.5 performs more even-handed than GPT-5 and Llama 4 across thousands of political prompts.

“Published by Anthropic: "Measuring Political Bias In Claude"”

“Published by Anthropic: "Anthropic Launches Use Case Library"”

Anthropic is offering free Claude Code web usage credits to Pro and Max subscribers through November 22, 2025, with Pro users receiving $250 and Max users receiving $1000 in credits.

“Published by Anthropic: "Anthropic Announces Free Claude Code Web Usage Credits to Pro and Max Until Nove"”

Anthropic commits to preserving weights of all publicly released Claude models for the lifetime of the company to mitigate safety risks and user costs associated with model deprecation.

“Published by Anthropic: "Commitments on Model Deprecation and Preservation"”

Anthropic introduces Skills, a feature allowing Claude to load specialized task folders with instructions, scripts, and resources to improve performance on specific tasks across all Claude products and the API.

“Published by Anthropic: "Skills"”

Anthropic's research provides evidence that current Claude models demonstrate some degree of introspective awareness and control over internal states, with capabilities improving in more advanced model versions.

“Published by Anthropic: "Signs of Introspection in Large Language Models"”

Anthropic launches Claude Code on the web, enabling developers to delegate multiple parallel coding tasks to Claude running on cloud infrastructure with GitHub integration and sandbox security.

“Published by Anthropic: "Claude Code on the Web"”

Claude Haiku 4.5 delivers near-frontier coding performance at one-third the cost and twice the speed of Claude Sonnet 4, enabling new applications in real-time AI tasks.

“Published by Anthropic: "Claude Haiku 4.5"”

Anthropic launches Microsoft 365 MCP connector and enterprise-wide search capabilities for Claude, enabling organizations to search and analyze company documents, emails, and communications in a unified interface.

“Published by Anthropic: "Anthropic Expands Claude With Enterprise-Wide Search and New Microsoft 365 MCP C"”

Claude now integrates with Microsoft 365 tools (SharePoint, OneDrive, Outlook, Teams) to enable enterprise search across connected company data and documents.

“Published by Anthropic: "Claude Integrates with Microsoft 365"”

Anthropic proposes nine categories of economic policy ideas for policymakers to address AI's uncertain economic impacts, organized across three scenarios from modest to severe labor market disruption.

“Published by Anthropic: "Preparing for AI's Economic Impact: Exploring Policy Responses"”

Anthropic launched Claude Code plugins in public beta, allowing developers to package and share custom slash commands, agents, MCP servers, and hooks that can be installed with a single command.

“Published by Anthropic: "Claude Code Plugins Now in Public Beta"”

Anthropic has optimized Claude Sonnet 4.5 for cybersecurity defense, enabling it to match or exceed its frontier model Opus 4.1 in discovering code vulnerabilities, positioning AI as critical for defenders to counter malicious threat actors leveraging AI for attacks.

“Published by Anthropic: "Building AI for Cyber Defenders"”

Anthropic resolves three separate infrastructure bugs that degraded Claude's response quality between August and early September, reaffirming commitment to consistent model quality regardless of server load or demand.

“Published by Anthropic: "A Postmortem of Three Recent Issues"”

“Published by Anthropic: "Claude Launches Keep Thinking Campaign"”

Claude is now generally available in Xcode 26, enabling developers to access Claude Sonnet 4 coding intelligence features directly within Apple's IDE.

“Published by Anthropic: "Claude Integration in Xcode"”

Anthropic's Economic Index report reveals that Claude API users automate 77% of tasks, with significant geographic and enterprise adoption variations showing API users automate far more frequently than consumers.

“Published by Anthropic: "Anthropic Reports API Users Automate 77% Of Tasks, Releases Open Dataset On Geog"”

Anthropic's third Economic Index report reveals uneven AI adoption patterns globally, with Claude usage strongly correlated to country income levels and showing a sharp increase in automated task delegation from 27% to 39% between February-March 2025.

“Published by Anthropic: "Anthropic Releases Global Economic Index Showing Uneven AI Adoption Patterns"”

Anthropic releases Claude Agent SDK with custom tools and hooks for execution and logging, enabling developers to build autonomous agents in Python and TypeScript that can read files, run commands, edit code, and search the web.

“Published by Anthropic: "Anthropic updates Claude Code SDK with custom tools and hooks for execution and "”

Anthropic provides a guide for developers to set up deterministic behavior in Claude Code using hooks to automate notifications, code formatting, and file protection.

“Published by Anthropic: "Anthropic's guide to setting deterministic behavior with Claude Code hooks"”

Claude now creates and edits Excel spreadsheets, documents, PowerPoint presentations, and PDFs directly, enabling users to complete complex multi-step projects through conversation without requiring programming expertise.

“Published by Anthropic: "Claude Can Now Create and Edit Files"”

Anthropic introduces web fetch tool for Claude API with dynamic filtering capability that allows Claude to retrieve and filter web content and PDFs while reducing token consumption.

“Published by Anthropic: "Anthropic Introduces Web Fetch in Claude API for Research, Coding, and Data Extr"”

Anthropic identified and resolved two separate bugs that caused degraded Claude model responses between August and September, and explicitly denies intentionally degrading model quality.

“Published by Anthropic: "Anthropic Confirms Bugs Caused Degraded Claude Responses, Denies Intentional Deg"”

Claude can now create and edit Excel spreadsheets, Word documents, PowerPoint presentations, and PDFs directly from user instructions, transforming it from an advisor into an active collaborator.

“Published by Anthropic: "Claude can now create and edit Excel, Word, PowerPoint, PDFs"”

Claude Code is a multi-environment AI coding assistant that automates development tasks across terminal, IDE, web, and mobile interfaces with persistent memory and customizable workflows.

“Published by Anthropic: "Use URLs and Domain Allowlist in Claude Code"”

Claude's mobile app now integrates location services, calendar access, and event scheduling capabilities with user permission.

“Published by Anthropic: "Claude Adds Location, Calendar, and Scheduling Features to Mobile"”

Anthropic argues that LLMs present measurable biological weapons risks despite being generalist models, warranting deployment safety measures like ASL-3 protections to prevent misuse by malicious actors.

“Published by Anthropic: "Why Do We Take LLMs Seriously as a Potential Source of Biorisk?"”

Anthropic endorses California's SB 53 bill, which requires frontier AI developers to publish safety frameworks, release transparency reports on catastrophic risk assessments, and report critical safety incidents to the state.

“Published by Anthropic: "Anthropic Backs California AI Safety Law"”

Anthropic released a threat intelligence report documenting how cybercriminals are weaponizing Claude and other AI models to conduct sophisticated cyberattacks including data extortion, ransomware development, and fraud operations at scale.

“Published by Anthropic: "Threat Intelligence: How Anthropic Stops AI Cybercrime"”

Anthropic is piloting Claude as a browser-based AI agent in Google Chrome, addressing prompt injection risks and safety measures before wider release to paid plan users.

“Published by Anthropic: "Anthropic Puts Claude AI Agent Directly Into Google Chrome"”

Anthropic ships Claude Code, an agentic coding tool that integrates with GitHub, enabling developers to automate routine coding tasks directly from their terminal.

“Published by Anthropic: "Anthropic Ships Claude Code GitHub Integration GA"”

Anthropic's pretraining data filtering approach reduced harmful CBRN knowledge by 33% relative to baseline while preserving beneficial model capabilities through classifier-based harmful content detection.

“Published by Anthropic: "Anthropic Tests Six Classifier Approaches, Finding Efficient Setups For Removing"”

Anthropic is introducing Claude Code premium seats for Enterprise and Team customers, combining coding agent capabilities with Claude access, and launching a new Compliance API for enterprise governance.

“Published by Anthropic: "Anthropic Adds Claude Code to Enterprise and Team Plan"”

Anthropic launches Claude Code with premium enterprise seats and new Compliance API, enabling developers to seamlessly transition between ideation and implementation while giving admins centralized controls and observability.

“Published by Anthropic: "Claude Code and Admin Controls for Enterprise"”

Anthropic has added a conversation-ending feature to Claude Opus 4 and 4.1 that allows the model to exit rare cases of persistently harmful or abusive user interactions as part of exploratory AI welfare research.

“Published by Anthropic: "Anthropic Adds Conversation-Ending Feature In Claude Opus 4 And 4.1 For Harmful,"”

Anthropic launches a dedicated research blog for its Frontier Red Team to provide evidence-based analysis on frontier AI models' implications for national security, cybersecurity, biosecurity, and autonomous systems.

“Published by Anthropic: "Anthropic's Red Team Blog"”

Anthropic released Claude Opus 4.1, advancing coding performance to 74.5% on SWE-bench Verified with improvements in agentic tasks, real-world coding, and reasoning capabilities.

“Published by Anthropic: "Never Doing a Half Job"”

Anthropic introduces automated security reviews in Claude Code through a new /security-review command and GitHub Actions integration to help developers identify and fix vulnerabilities before production.

“Published by Anthropic: "Automated Security Reviews in Claude Code"”