In its Responsible AI Transparency Report, which mainly covers 2023, Microsoft touts its achievements around safely deploying AI products. The annual AI transparency report is one of the commitments the company made after signing a voluntary agreement with the White House in July last year. Microsoft and other companies promised to establish responsible AI systems and commit to safety.

Microsoft says in the report that it created 30 responsible AI tools in the past year, grew its responsible AI team, and required teams making generative AI applications to measure and map risks throughout the development cycle. The company notes that it added Content Credentials to its image generation platforms, which puts a watermark on a photo, tagging it as made by an AI model. 

The company says it’s given Azure AI customers access to tools that detect problematic content like hate speech, sexual content, and self-harm, as well as tools to evaluate security risks. This includes new jailbreak detection methods, which were expanded in March this year to include indirect prompt injections where the malicious instructions are part of data ingested by the AI model.

It’s also expanding its red-teaming efforts, including both in-house red teams that deliberately try to bypass safety features in its AI models as well as red-teaming applications to allow third-party testing before releasing new models.

However, its red-teaming units have their work cut out for them. The company’s AI rollouts have not been immune to controversies.

Natasha Crampton, chief responsible AI officer at Microsoft, says in an email sent to The Verge that the company understands AI is still a work in progress and so is responsible AI. 

“Responsible AI has no finish line, so we’ll never consider our work under the Voluntary AI commitments done. But we have made strong progress since signing them and look forward to building on our momentum this year,” Crampton says. 

Shares:

Leave a Reply

Your email address will not be published. Required fields are marked *