- Jenny Lay-Flurrie, leader of Microsoft's Trusted Technology Group, states that ethical technology involves questioning how we construct it properly and how we maintain its integrity.
- In early 2025, Microsoft introduced its Trusted Technology Group and has since brought all ethical technology efforts under one roof, covering areas such as accessibility, digital security, human rights, ethical AI, privacy, supply chain reliability, and technology aimed at societal benefit.
- When Microsoft recognized that its AI was not properly depicting blind individuals, Lay-Flurrie's group took action to resolve the issue.
Completely reliable and dependable technology is nearly unattainable in a tech environment that values rapid development — yet this doesn't stop certain companies from making an effort.
Following the Trump administration'snational AI legislative frameworkOn March 20, where "winning the AI race" continues to be a top priority, technology creators encounter conflict between the traditional mindset of rapid development and the need to incorporate ethical tech guidelines from the beginning.
In many cases, moving forward has taken precedence, with the consequences becoming evident. Microsoft's own acknowledgment thatAI-generated codeoften skips accessibility, making human review and refinement essential.
For Jenny Lay-Flurrie, who took on the role of head of Microsoft's Trusted Technology Group in February and has dedicated much of her 21-year career at the company to accessibility, the ethical creation and implementation of technology involves two key aspects: "How do we ensure that we get it right in the first place? And how can we keep it right over time?"
Microsoft introduced its Trusted Technology Group in early 2025 and has now brought all responsible technology efforts under this organization, including the previous initiative on accessibility led by Lay-Flurrie.
While Microsoft has consolidated its responsible technology under a top-down structure, rivals such as Google adopt a more engineering-focused framework influenced by its fundamental AI values and dedicated safety committees. Methods differ among major tech companies, but Microsoft's strategy has evolved since 2002, when Bill Gates published theTrustworthy Computing memothat focused on reliability rather than introducing new features.
The challenges of AI (and those who resolve them)
Lay-Flurrie's entry into the wider field of responsible technology might be new, but she states that it aligns with the same core values she has always applied, such as fairness, transparency, inclusiveness, and accountability.Microsoft functions based on the principlethat "individuals must take responsibility for artificial intelligence" irrespective of the results.
That's why, when Microsoft recognized that its AI was not properly portraying individuals who are blind, her team took action to resolve the issue.
Some of the images generated of blind individuals ended up showing people wearing intense, full-face blindfolds," she explained. "These models are trained on a vast amount of existing content. Regrettably, society isn't always the most inclusive environment, so there are times when we need to add specific data to improve the training.
To accomplish this, Microsoft acquired over 20 million minutes of multimodal data from Be My Eyes, a nonprofit accessibility platform that is free for blind and low-vision individuals to use in order to connect with live volunteers and AI, providing them with audio descriptions of what they are seeing. "They had a significant amount of video content captured by blind individuals showing their use of canes and guide dogs, as well as locating keys within their homes, and we anonymized the data by blurring faces and other identifying features so that we could better train our models for blindness," explained Lay-Flurrie.
This method is reliable, but Annie Brown, CEO and founder of Reliabl, a machine learning training platform aimed at reducing bias and enhancing AI model effectiveness, mentioned there is still potential for enhancement.
Having more varied data is only one aspect," noted Brown. "If you don't focus on what's occurring at the metadata level, which involves how the images uploaded to your dataset are categorized, this alone can lead to bias.
Even though the AI competition is reshaping the world, Microsoft is involved in a larger trend where companies openly share their insights on responsible technology. Microsoft Learn is accessible at no cost to students, researchers, and developers, and offers training modules onresponsible AI principlesand more. Brown suggests that Microsoft should also learn from smaller social good organizations to understand "how they are integrating inclusivity into AI."
Regarding improvements, Lay-Flurrie mentions that it's part of the process. "It involves clearly listening to feedback, accepting it, making changes, testing, and addressing issues as quickly as possible," she stated.
Humans vs. AI
Microsoft is a leading supplier of enterprise technology, which means its own AI is driving other companies to often replace workers with more advanced solutions. Microsoft is also part of a larger trend of job cuts in big tech, although it has stated that this is more about shifting focus than simply replacing employees. The companycut roughly 15,000 jobsIn 2025, there was an emphasis on sales, gaming, and customer-facing departments, with new hires brought on in other areas focusing on AI infrastructure.
Even with ongoing job cuts in various sectors, Lay-Flurrie claims AI is already creating a more balanced environment for workers who have historically been overlooked, including those who areneurodiverse and disabled.
The initial group to receive access to Copilot at Microsoft was our disability employee community," she stated. "For the Deaf community, features like captions, transcripts, meeting summaries, and sign language recognition offer autonomy. You don't need to wait for a transcriptionist to be present to capture what is being said.
For the neurodiverse community who had access to Copilot early, it significantly reduced the mental burden, she mentioned, saying "they wouldn't let me return the license."
Diego Mariscal, the CEO and founder of the global startup accelerator 2Gether-International (2GI), which is operated by and for entrepreneurs with disabilities, acknowledges that Microsoft has made a commitment to involve individuals with disabilities. "The mere existence of Jenny's role at this level is proof of that," he mentioned. Nevertheless, it's crucial to have people with disabilities involved in decision-making processes, both from the top down and the grassroots up. "How can we make sure that as AI develops, individuals with disabilities are part of the conversation, not just from a charitable viewpoint, but because their inclusion will lead to technology and innovation that is more advanced and accessible for all?"
No comments:
Post a Comment