Microsoft has announced a new multiyear partnership with Mistral, a French AI startup, to bring Mistral’s language models to Microsoft’s Azure AI platform. This partnership follows Microsoft’s previous investment in OpenAI and will focus on developing and deploying next-generation large language models. Mistral has also introduced Mistral Large, a new AI model designed to compete with OpenAI’s GPT-4 model. The details of the investment have not been disclosed.
GitHub has launched Copilot Enterprise, an AI coding assistant aimed at helping large enterprises with their coding projects. Copilot Enterprise integrates throughout the software development lifecycle, providing customized code suggestions, answering queries, and summarizing code changes. It incorporates company best practices and documentation, enabling developers to share collective knowledge and enforce organizational standards.
Microsoft has announced “AI Access Principles” to address concerns about competition with OpenAI. The eleven-point plan includes commitments to allow businesses to choose different AI products, keep proprietary data out of training models, and enable customers to change cloud providers. The announcement comes as Microsoft faces regulatory scrutiny for its investment in OpenAI. While the principles are not binding, they serve as a public reference point for measuring Microsoft’s actions.
Microsoft has also released its Python Risk Identification Toolkit for generative AI, or PyRIT, to help identify risks in generative AI systems. The toolkit, used by Microsoft’s AI Red Team, automates the probing process for security and responsible AI risks, making red teaming more efficient and time-saving. PyRIT sends a malicious prompt to the generative AI system, scores the response, and sends a new prompt based on the feedback. The toolkit is available for access and includes demos to familiarize users with the tool.
IT security teams and developers should incorporate PyRIT into their security practices for AI systems, and developers should consider tools like CoPilot Enterprise. These tools can help identify vulnerabilities early in development, accelerate development timelines, improve code quality, and ensure that AI applications are secure and aligned with ethical guidelines.
Microsoft is trying to be multi-platform and benefit from investing in a single player. That’s the strategy to note – and it potentially helps protect against anti-trust.