With cloud giants competing for AI application opportunities, Amazon's AWS, which focuses on corporate applications, emphasizes its security advantages.
There have been concerns about AWS not being as active as Microsoft in its approach to generative AI. However, the Microsoft-backed OpenAI continues to face controversies, including data leaks, copyright issues with training data, and a recent case where it may have used the voice of famous actress Scarlett Johansson without her consent.
These controversies present an opportunity for AWS to enhance its strengths. Anthropic, an AI startup heavily invested in by Amazon, is known for its focus on secure applications.
AWS stated that for businesses to use generative AI safely, three key issues should be considered: the location of the data; the handling of input and output data during model inference; and the accuracy of the generative AI model's output. Different usage scenarios have varying requirements for accuracy and risk.
AWS emphasized that its unique cloud server virtualization engine, AWS Nitro, can download the Input/Output (I/O) function from the host CPU/GPU to specialized hardware, thus protecting the user's code and data processing on both the client and cloud sides.
The Nitro system provides hardware-level security mechanisms, and its end-to-end encryption processes can extend to AWS Trainium 2 chips and other GPUs, ensuring secure data communication between infrastructure equipment.
The Amazon Bedrock host service platform offers Guardrails for Amazon Bedrock, a responsible AI protection mechanism. Unlike some AI models that filter content solely through internal control modules, this feature allows customers to further customize AI applications to meet different content policy standards.
The chatbot development assistant Amazon Q Developer has built-in security scanning features. It is capable of detecting hard-to-discover loopholes and providing specific repair recommendations.
AWS stated that generative AI will become as ubiquitous as the internet and cloud computing are today. While new digital technologies will inevitably bring new security challenges, generative AI technologies will also aid security innovation, making cybersecurity features simpler and easier to use.
Article translated by Jack Wu
