By 
In the wake of Microsoft’s recent investment and collaboration with French Large Language Model startup Mistral AI, the company is actively working to dispel any perception of hindering competition through its close partnership and financial involvement with OpenAI.
Microsoft this week introduced a new framework termed “AI Access Principles” — an eleven-point agenda that the company asserts will “guide our operations concerning AI datacenter infrastructure and other significant AI assets worldwide.”
These principles encompass various aspects, including the establishment and management of an app store facilitating businesses to select from a range of different LLMs and other AI products. Additionally, Microsoft commits to excluding its proprietary data from its training models, allowing customers the freedom to switch cloud providers or services within the cloud at their discretion. The framework also emphasizes cybersecurity measures around AI services, environmentally sustainable infrastructure development, and investments in education.
Brad Smith, President and Vice Chair of Microsoft, unveiled the framework at the Mobile World Congress in Barcelona today. While the announcement signals Microsoft’s willingness to engage in dialogue with stakeholders, it is notable that Smith delivered the news in a keynote speech without the opportunity for follow-up questions.
This announcement coincides with increasing regulatory scrutiny faced by Microsoft due to its $13 billion investment in OpenAI, granting it a 49% stake in the leading generative AI services startup globally. In January, the European competition watchdog announced an assessment of whether the investment violates antitrust rules.
The principles particularly address how third parties may leverage Microsoft’s platforms and services to develop AI products, a crucial sector that Microsoft aims to expand in the coming years. Smith emphasized, “If they are training a model on our infrastructure, if they are deploying it on our infrastructure, we recognize that their data is their data; we will not access it to compete with the companies relying on our infrastructure.”
It’s crucial to note that these AI Access Principles are not legally binding rules for Microsoft, nor are there specifics outlined regarding how these commitments will be verified or monitored. However, they serve as a preemptive measure for potential regulatory inquiries. In the event of formal investigations, Microsoft could leverage these principles to demonstrate proactive efforts
towards fostering competition in the market.
Smith highlighted Microsoft’s current utilization of nearly 1600 models in its data centers, with 1500 of them being open-source models, underscoring the company’s commitment to both proprietary and open-source models, as well as enterprises of all sizes.
By publicly articulating these principles, Microsoft aims to provide a reference point for the public, competitors, and regulators to gauge its adherence to fostering a competitive AI landscape.
