Connecting the dots to AI Security in the Cloud—Thoughts from AWS reInvent
Dots and Boxes is a pencil and paper game for two players (sometimes more). It was first published in the 19th century by French mathematician Edouard Lucas, who called it la pipopiette. Players take turns adding a single horizontal or vertical line between unjoined, adjacent dots.
According to Statista, Amazon AWS (cloud) is the most successful business segment within Amazon, and it generated over $80 billion in revenues in 2022 and almost $23 billion in operating profit.
The largest of all the hyperscalers, AWS has placed big bets on AI, investing up to $4 billion in AI startup Anthropic making its future foundation models accessible to millions of developers and providing AWS customers early access to unique features for model customization, using their proprietary data, and fine-tuning capabilities, all through Amazon Bedrock.
AWS is busy connecting the Cloud and AI dots.
AWS—All in on AI and AI Security
Ten days ago at reInvent, myself and about 60,000 others gathered to learn and listen and AI was THE topic of conversation
AWS invited me as an analyst to the event and I got a front row seat to a number of significant announcements in the space that ranged from more LLM choice, to support of multi-modal vector embeddings to Amazon Q and model evaluation on Amazon Bedrock in preview
Because Cloud Security is my jam, I was most interested in Guardrails for Amazon Bedrock (Preview)
Amazon Bedrock is a fully managed service that makes leading foundation models from AI companies available through an API along with developer tooling to help build and scale generative AI applications
In case you missed this security announcement, Guardrails for Amazon Bedrock brings a consistent level of AI safety across all your applications
Guardrails for Amazon Bedrock evaluates user inputs and Foundational Model (FM) responses based on use case specific policies, and provides an additional layer of safeguards regardless of the underlying FM.
Guardrails can be applied across FMs, including Anthropic Claude, Meta Llama 2, Cohere Command, AI21 Labs Jurassic, and Amazon Titan Text, as well as fine-tuned models. Customers can create multiple guardrails, each configured with a different combination of controls, and use these guardrails across different applications and use cases. Guardrails can also be integrated with Agents for Amazon Bedrock to build generative AI applications aligned with your responsible AI policies.
Business Value of AI Security
Guardrails for Bedrock helps customers to create a security framework for different use cases and apply them across multiple FMs. The goal is to provide a consistent user experience and standardize safety controls across generative AI applications.
While many FMs doe have built in protections to filter undesirable and harmful content, Guardrails allows customers to take things a step further and tailor interactions specific to their use cases. Customers can define a set of denied topics within the context of their application and configure thresholds to filter harmful content across categories. Guardrails then evaluates user queries and FM responses against the denied topics and content filters, helping to prevent content that falls into restricted categories. This allows customers to closely manage user experiences based on application specific requirements and policies. To me this a working template that will enable customers to create AI policies or bolster them.
AI Security–Building on a Cloud Security Foundation
Amazon has a full suite of cloud security offerings that have grown and matured over time. There are a total of 32 tools across 5 areas—Identity and Access Management, Detection and Response, Network and Application Protection, Data Protection and Compliance.
Guardrails for Bedrock is a foundational AI security tool. Organizations are looking for thought leadership in the quickly evolving AI space. Look to AWS to continue to connect the AI security dots in this growing space.
#cloud #cloudsecurity #aisecurity