Making Sense Of Tech Companies’ AI Commitments To The White House

On July 21, 2023, the White House announced that seven prominent tech companies, including Amazon, Google, ...

– Meta, Microsoft, and AI-focused startups Anthropic, Inflection, and OpenAI, made voluntary commitments concerning their AI systems.

The commitments are categorized into three groups: (1) safety, (2) security, and (3) trust.

– However, these commitments are not binding, and there is no enforcement mechanism if the companies fail to comply.

Some commitments, such as internal and external security testing of AI systems before release, are sensible and likely already practiced by the companies.

– The use of external parties for testing raises questions about the testing process, certification, and duration.

– Concerns arise regarding the competitive disadvantage faced by companies adhering to commitments while others not signing on can swiftly bring their AI products to market without third-party testing delays.

The fifth commitment involves developing technical mechanisms, like watermarking, to identify AI-generated content.

– Google has already implemented such features, and it aligns with proposed legislation for disclosure of AI use in political ads.

Other commitments, like using advanced AI systems to address societal challenges, are seen as aspirational goals

– leaving room for companies to pursue other objectives, potentially less noble in nature.

The White House's approach signifies a willingness to collaborate with companies on AI-related policies

– though it may be viewed as toothless by those concerned about AI's social impacts. This cautious and measured approach aims to prevent policy mistakes that could hinder innovation.