News

Microsoft Announces 11 AI Access Principles

• By Kurt Mackie
• February 27, 2024

Microsoft described 11 corporate "AI access principles," aimed at fostering an artificial intelligence (AI)-based economy.

Brad Smith, Microsoft's vice chair and president, explained the voluntary principles in a Monday announcement. Smith depicted Microsoft's role in building an "AI economy" as being largely benevolent, rather than monopolistic, with high dependence on partners. It included assurances on that theme for government regulators. AI regulation is currently under consideration via the European Union's Digital Markets Act and AI Act, and "the United States is moving quickly with a new AI Executive Order," the announcement noted.

Microsoft indicated that it is planning to "engage constructively with regulators when obligations are not yet clear" concerning AI, which likely offers the company some wiggle room.

Partnerships Needed
The future AI economy will depend quite a lot on Microsoft's partnerships with governments, the public and business partners, the announcement indicated.

"We are building the infrastructure that will support the AI economy, and we need the opportunities provided by that infrastructure to be widely available," the announcement asserted.

In terms of direct AI technology partnerships, Microsoft pointed to its long-time partnership with generative AI company OpenAI. Also, it announced on Monday that it was newly partnering with France-based Mistral AI. Microsoft is offering its Azure AI infrastructure and supercomputing capabilities for use with Mistral's next-generation large language models.

Microsoft's 11 AI Access Principles
Microsoft's 11 AI access principles are encompassed by three broad themes. Microsoft intends to support AI developers, for instance. It pledged to observe "fairness across the AI economy." Lastly, Microsoft declared that it would strive to "meet our societal responsibilities."

Those three broad themes go beyond Microsoft's U.S. corporate charter requirements, which is simply to achieve a profit. However, the announcement was less of a legal document than a public policy declaration. It seemed more aspirational than legally binding.

In a nutshell, here are Microsoft's 11 AI access principles, mostly aimed at developers using Azure services. Microsoft pledged to do the following things:

1. Expand its "cloud computing AI infrastructure" to support AI models, big and small. Microsoft currently hosts "more than 1,600 open source and proprietary models."
2. Make its "AI models and development tools broadly available" to developers around the world, typically via its GitHub repository. To that end, Microsoft recently released its PyRIT open source tool for red team security probes of AI models.
3. Make its application programming interfaces (APIs) publicly available to developers using the Azure OpenAI service so that they can access its "underlying models."
4. Provide support for a "common public API through the Open Gateway initiative driven by the GSM Association" to support network operators with their AI software developments. Microsoft also announced the release of a public preview of Azure Programmable Connectivity, which is an Azure service that "seamlessly provides access to Open Gateway for developers."
5. Permit developers using Azure infrastructure to "distribute and sell their AI models, tools and applications" not just through the Azure Marketplace. They can also distribute and sell their products "directly to customers."
6. Pledged to not use developer data or models to train Microsoft's AI models.
7. Pledged that AI developers can switch to another cloud service provider by exporting their data. Microsoft also promised to follow laws that might limit its fees associated with such data exports.
8. Provide physical datacenter security, including encryption to protect data "in transit and at rest." These efforts are based on Microsoft's "Secure Future Initiative," announced last year.
9. Follow the "Microsoft Responsible AI Standard" for creating "ethical" and "trustworthy" AI technologies.
10. Invest in AI skills around the world via Microsoft certified trainers, industry partners, nonprofit organizations and universities. Microsoft touted its AI partnership with the American Federation of Labor and Congress of Industrial Organizations (AFL-CIO) labor union as one example.
11. Manage datacenters in an "environmentally sensitive manner." However, one countercurrent to Microsoft's view is the notion that the widespread AI use will increase energy consumption to a large degree, per a recent estimate in the journal Joule.

The 11 principles were described as being "only a first step." Smith added that "we expect that we will need to evolve these principles and our approach as AI technology and industry moves forward and the applicable law and regulations change."