Amazon has joined forces with President Biden and leaders from government and industry to pledge their commitment to promoting the safe, secure, and transparent development of artificial intelligence (AI) technology that benefits society. The White House convened a meeting with leading AI companies to identify commitments related to the responsible development of future generative AI models.
These commitments build on Amazon’s existing approach to responsible and secure AI development, and they aim to pave the way for a future that maximizes the benefits of AI while minimizing its risks. Amazon has made a series of voluntary commitments to support these goals.
Firstly, the company has committed to conducting internal and external adversarial-style testing, commonly known as “red-teaming,” of AI models or systems. This testing will focus on areas such as misuse, societal risks, and national security concerns to ensure potential vulnerabilities are identified and addressed.
Secondly, Amazon aims to facilitate information sharing among companies and governments regarding trust and safety risks associated with AI. This includes sharing knowledge about dangerous or emergent capabilities and any attempts to circumvent safeguards that are in place.
In addition, Amazon is dedicated to developing and deploying mechanisms that allow users to determine whether audio or visual content has been generated by AI. This includes implementing robust provenance, watermarking, or both, for AI-generated audio or visual content.
Cybersecurity and insider threat safeguards are also a priority for Amazon. The company will invest in measures to protect proprietary and unreleased model weights from unauthorized access or compromise.
Furthermore, Amazon will incentivize third-party discovery and reporting of issues and vulnerabilities to ensure that potential risks are identified and addressed promptly. The company will report publicly on the capabilities, limitations, and appropriate use of its AI models or systems, including discussing societal risks such as fairness, bias, and privacy concerns.
Amazon recognizes the importance of researching and addressing societal risks posed by AI systems, including issues related to harmful bias, discrimination, and privacy infringement. The company is dedicated to developing and deploying advanced AI systems to tackle some of society’s greatest challenges.
Amazon is committed to collaborating with the White House, policymakers, the technology industry, researchers, and the AI community to promote the responsible and secure use of AI technology. As one of the world’s leading developers and deployers of AI tools and services, Amazon fully supports the voluntary commitments outlined to foster the development of safe, responsible, and trustworthy AI.
Amazon’s approach to responsible AI development aligns with the commitments made by the White House. Throughout the comprehensive development process, Amazon prioritizes factors such as accuracy, fairness, appropriate usage, toxicity, security, safety, and privacy. The company continuously improves its AI features based on customer feedback and encourages experimentation with new use cases.
Responsible use of AI technology is crucial for ongoing innovation, and Amazon is dedicated to providing fair and accurate AI services. For example, Amazon CodeWhisperer incorporates built-in security scanning to identify and suggest fixes for hard-to-detect vulnerabilities. It also offers source code provenance analysis and indicates the applicable open source licensing models for generated code.
Additionally, Amazon’s Titan family of foundation models is designed to detect and remove harmful content, reject inappropriate user input, and filter model outputs containing hate speech, profanity, or violence. The company’s SageMaker Clarify tool detects and measures potential bias, allowing developers to address bias and explain model predictions.
To empower customers to develop and use AI responsibly, Amazon has launched AWS AI Service Cards. These cards serve as a transparency resource, providing customers with information on use cases, limitations, responsible AI design choices, and best practices for deploying and optimizing AI services.
Security is a paramount concern for Amazon in AI development. The company rigorously tests all its AI services for security and maintains strict physical, electronic, and procedural safeguards. Amazon actively collaborates with the security community to strengthen its products and systems and provides channels for reporting vulnerabilities.
As a transformational technology, AI holds immense potential for positive impact. Amazon is committed to working with customers and partners to innovate and harness the benefits of AI while upholding shared commitments to responsible development.
link