President Biden has secured voluntary commitments from seven leading AI companies who say they will support the executive branch goal of advancing safe, secure and transparent development of artificial intelligence. Executives from Amazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI convened at the White House on Friday to support the accord, which some criticized as a half measure, claiming the companies have already embraced independent security testing and a commitment to collaborating with each other and the government. Biden stressed the need to deploy AI altruistically, “to help address society’s greatest challenges.”

The New York Times calls it “an early, tentative step as Washington and governments across the world rush to put in place legal and regulatory frameworks for the development of artificial intelligence.”

It can also be seen as the corporate corollary to the administration’s AI Bill of Rights for consumers, and another step in advancing national best practices with regard to rapidly evolving AI.

NYT points out that “as voluntary commitments, they won’t be enforced by government regulators.” But they amplify the White House messaging that AI is a priority, one the administration is going to try to get out ahead of, unlike social media.

Among the protocols the seven companies agreed to:

• Ensuring product safety prior to commercialization — including security testing for biosecurity, cyber threats and other vulnerabilities.
• Sharing best practices with each other, with government and academia. This includes technical collaboration, as well as “information on attempts to circumvent safeguards,” according to a White House fact sheet.
• Protecting “proprietary and unreleased model weights” — Weights and biases are neural network parameters that determine how data flows through the network (also called “forward propagation”).
• Post-release reporting mechanism to share information about vulnerabilities, allowing them “to be found and fixed quickly.”
• Earning Public Trust through things like watermarking that indicates when content has been artificially generated, and sharing information on AI biases as a step toward reducing it.

Biden, who gave a brief address summarizing the pact, offered no further details on what NYT described as “a forthcoming presidential executive order that will deal with a bigger problem: How to control the ability of China and other competitors to get a hold of the new artificial intelligence programs, or the components used to develop them.”

The upcoming executive order is expected to place limits on the export of large language models, which NYT points out “are hard to control,” with much of the software able to fit, compressed, onto a thumb drive. The Verge says the executive order could play out across multiple federal agencies.

Related:
Top Tech Firms Sign White House Pledge to Identify AI-Generated Images, Washington Post, 7/21/23
OpenAI Backs Idea of Requiring Licenses for Advanced AI Systems, Bloomberg, 7/20/23
OpenAI, Google, Others Pledge to Watermark AI Content for Safety, Reuters, 7/21/23
How Do the White House’s AI Commitments Stack Up?, The New York Times, 7/22/23
Leading AI Companies Commit to Outside Testing of AI Systems, CNN, 7/21/23