On September 25, 2023, Amazon announced it will invest up to $4 billion in Anthropic and enter a broader collaboration to develop and deploy reliable, high-performance foundation models with an emphasis on safety.
Scope of the collaboration
-
Amazon Web Services (AWS) will become Anthropic’s primary cloud provider for mission-critical workloads. Anthropic will gain access to AWS Trainium and Inferentia chips to supplement its existing infrastructure for model training and deployment. The companies will also collaborate on the development of future Trainium and Inferentia technologies.
-
Support for Amazon Bedrock will be expanded to include secure model customization and fine-tuning of Anthropic models, enabling enterprises to optimize Claude’s performance with their domain expertise while reducing the risk of harmful outcomes.
-
Amazon developers and engineers will be able to build on Anthropic’s state-of-the-art models via Amazon Bedrock, integrating generative AI capabilities into existing applications or creating new customer experiences across Amazon’s businesses.
Claude 2 and technical features
Anthropic’s Claude 2 is positioned for a wide range of tasks, from sophisticated dialogue and creative content generation to complex reasoning and detailed instruction-following. The announcement highlights Claude 2’s industry-leading 100,000 token context window, which allows secure processing of large volumes of information, including technical and domain-specific documents relevant to finance, legal, and coding use cases.
Customer use cases
The release cites several organizations already building with Anthropic models on Amazon Bedrock:
-
LexisNexis Legal & Professional uses a custom, fine-tuned Claude 2 model to provide conversational search, summarization, and intelligent legal drafting as part of the Lexis+ AI solution.
-
Bridgewater Associates is developing an investment analyst assistant powered by Claude 2 to generate complex charts, compute financial indicators, and produce summaries.
-
Lonely Planet reduced itinerary-generation costs by nearly 80 percent after deploying Claude 2, using the model to synthesize decades of travel content into coherent, accurate recommendations.
Safety and responsible development commitments
Both Anthropic and Amazon state commitments to the safe training and deployment of advanced foundation models. Amazon, as a market leader in cloud security, has committed under the agreement to promote and implement safety best practices on Amazon Bedrock to ensure responsible use of these products and services.
Both companies are active in organizations promoting responsible AI development and deployment, including the Global Partnership on AI (GPAI), the Partnership on AI (PAI), and the National Institute of Standards and Technology (NIST). In July, Amazon and Anthropic each supported a set of voluntary safety commitments led by the White House aimed at guiding the future of transformative AI with safety, security, and trust.
Ownership, governance and testing
As part of the investment, Amazon will take a minority stake in Anthropic. Anthropic’s corporate governance structure remains unchanged: the Long Term Benefit Trust will continue to guide the company under Anthropic’s Responsible Scaling Policy. That policy specifies that the company will carry out pre-deployment tests of new models to help manage risks as AI capabilities increase.
Resources and next steps
Training state-of-the-art models requires substantial resources, including compute power and research programs. Amazon’s investment and the provision of AWS Trainium and Inferentia technology are intended to equip Anthropic to continue advancing AI safety and research. The companies said they look forward to working closely to responsibly scale the adoption of Claude and deliver safe AI cloud technologies to organizations worldwide.
