February 2025 Summaries

Anthropic has partnered with the U.S. Department of Energy's National Laboratories to participate in the first-ever 1,000 Scientist AI Jam event. This session will bring together scientists from multiple laboratories to evaluate the capabilities of Claude 3.7 Sonnet, a hybrid reasoning model, on scientific research and national security applications. The AI Jam represents a significant milestone in Anthropic's partnership with DOE National Laboratories, building on previous collaborations such as the evaluation of Claude in a Top Secret classified environment by NNSA and DOE National Laboratories. This event aims to harness AI's transformative potential while addressing potential risks through rigorous testing, ultimately contributing to accelerating American scientific leadership and progress in solving humanity's greatest challenges.

Anthropic is launching its Transparency Hub, a detailed overview of measures being implemented to ensure the safety and trustworthiness of its AI systems. The hub provides information on methodologies for model evaluation and safety testing, platform abuse detection, internal governance policies, research initiatives, security safeguards, and more. It aims to address diverse documentation requirements across transparency initiatives and voluntary commitments, offering a unified framework for users, policymakers, and stakeholders to access clear insights into Anthropic's AI development and deployment. The hub is an ongoing commitment to raising the bar on transparency, with continuous expansion of reporting as AI capabilities advance and new challenges emerge.

Today we're announcing that Claude models are helping power Alexa+, a collaboration between Anthropic and Amazon to deliver advanced AI technology to businesses and consumers worldwide. Teams from both companies worked closely throughout the past year, with Anthropic's Chief Product Officer Mike Krieger leading a team that helped unlock Claude's capabilities for Alexa+. The partnership ensures that Alexa+ benefits from Anthropic's best-in-class safety features, including jailbreaking resistance. The integration will start rolling out in the U.S. in the next few weeks and become more widely available in the coming months.

Claude 3.7 Sonnet is the latest model from Anthropic, a hybrid reasoning model that can produce near-instant responses or extended, step-by-step thinking, with fine-grained control over how long the model thinks for. This model has shown particularly strong improvements in coding and front-end web development capabilities. Claude 3.7 Sonnet is now available on all plans, including Free, Pro, Team, and Enterprise, as well as on various APIs such as Anthropic API, Amazon Bedrock, and Google Cloud's Vertex AI. The model has the same price as its predecessors, with $3 per million input tokens and $15 per million output tokens. Claude 3.7 Sonnet embodies a unified approach to reasoning, where it can represent an ordinary LLM and a reasoning model in one, allowing users to pick when they want the model to answer normally or think longer before answering. The model also allows users to control the budget for thinking through the API, enabling them to trade off speed and cost for quality of answer. Additionally, Claude 3.7 Sonnet has improved its coding capabilities, including handling complex codebases, advanced tool use, and planning code changes. Claude Code is a new agentic coding tool that enables developers to delegate substantial engineering tasks to Claude directly from their terminal, reducing development time and overhead. Claude Code is available as a limited research preview and will be continually improved based on user feedback. The model has also been tested for security, safety, and reliability, with improvements in distinguishing between harmful and benign requests, and reducing unnecessary refusals by 45%.

Anthropic has signed a Memorandum of Understanding (MOU) with the UK's Department for Science, Innovation and Technology to explore how Artificial Intelligence can transform public services in the UK. The collaboration will focus on enhancing digital services, establishing best practices for AI deployment, and advancing scientific progress. Anthropic's advanced AI model, Claude, is expected to play a key role in this effort. The partnership aims to support AI innovation and economic development in the UK, with potential applications including securing supply chains and boosting the innovation economy. Additionally, the collaboration will draw on Anthropic's Economic Index to provide insights for adapting workforce and innovation strategies to an AI-enabled future, and work closely with the UK AI Security Institute to develop robust safeguards for responsible AI deployment.

We were pleased to attend the AI Action Summit in Paris, and we appreciate the French government’s efforts to bring together AI companies, researchers, and policymakers from across the world. We share the goal of responsibly advancing AI for the benefit of humanity. However, greater focus and urgency is needed on several topics given the pace at which the technology is progressing. The need for democracies to keep the lead, the risks of AI, and the economic transitions that are fast approaching—these should all be central features of the next summit. We must accelerate our actions to match accelerating AI progress. Possibly by 2026 or 2027 (and almost certainly no later than 2030), the capabilities of AI systems will be best thought of as akin to an entirely new state populated by highly intelligent people appearing on the global stage—a “country of geniuses in a datacenter”—with the profound economic, societal, and security implications that would bring. Democracies must lead in AI to prevent authoritarian countries from establishing global military dominance through its misuse. International conversations on AI must address the technology’s growing security risks, including catastrophic misuse risks or “loss of control” scenarios. Governments need to enforce transparency of safety and security plans, facilitate measurement of cyber attacks, CBRN, autonomy, and other global security risks, and enact policy focused on ensuring everyone shares in the economic benefits of very powerful AI. The advance of AI presents major new global challenges that require faster and greater clarity to confront them.

The Anthropic Economic Index is an initiative aimed at understanding AI's effects on labor markets and the economy over time. The index provides first-of-its-kind data and analysis based on millions of anonymized conversations on Claude.ai, revealing how AI is being incorporated into real-world tasks across the modern economy. The initial report shows that AI use is concentrated in software development and technical writing tasks, with around 36% of occupations using AI for at least a quarter of their associated tasks. AI leans more towards augmentation (57%) than automation (43%), with mid-to-high wage occupations like computer programmers and data scientists being the heaviest users of AI. The index also provides insights into how AI is changing the labor market, including the fact that very few occupations see AI use across most of their associated tasks, but more moderate use of AI is widespread. The dataset used for this analysis will be openly sourced to allow researchers to build on and extend the findings.

Lyft is collaborating with Anthropic to introduce AI-powered products that will enhance the rideshare experience for its community of over 40 million annual riders and over 1 million drivers. The collaboration focuses on three key areas: AI-Powered Solutions, Early Testing, and Engineering Advancement. The impact is already evident as Lyft's customer care AI assistant, powered by Claude via Amazon Bedrock, has reduced customer service resolution time by 87%. The partnership aims to make every interaction more efficient and personal for millions of people who rely on Lyft every day, and will deliver tangible benefits to its community today.