Q&A

Responsible AI Frameworks: An Introduction

• By David Ramel
• 05/22/2025

As the capabilities of AI continue to evolve at breakneck speed, so too does the need for clear ethical guardrails to guide its development and deployment. From bias mitigation to data provenance to ensuring transparency, the call for "responsible AI" has shifted from aspirational ideal to practical necessity -- particularly in light of today's generative models and enterprise-grade large language modesl (LLMs).

In response to this growing demand for ethical AI governance, numerous governments, organizations, and coalitions have released frameworks aimed at helping teams evaluate and improve the trustworthiness of their AI systems. But with so many guidelines -- ranging from the European Union's Ethics Guidelines for Trustworthy AI to tools developed by the OECD, Canada, and others -- it can be difficult for developers and decision-makers to know where to start or how to apply these frameworks in real-world projects.

That's where Karen Lopez comes in. A long-time data governance expert, space enthusiast, and NASA Datanaut, Lopez has spent the past several years studying publicly available Responsible AI frameworks, comparing their approaches, and identifying the most practical, actionable takeaways for enterprise teams. In her upcoming session, "Responsible AI Frameworks: An Introduction," at the San Diego 2025 edition of Visual Studio Live! on Sept. 10, Lopez will walk attendees through the ethical guidance that underpins responsible AI development, with a special focus on LLMs.

In the Q&A below, we spoke with Lopez about what developers can expect from her talk, how to apply frameworks like the EU guidelines to mitigate bias and hallucinations in LLMs, and what resources teams can use to start building more trustworthy AI -- whether or not they have a compliance team on staff.

VisualStudioMagazine: What inspired you to present a session on this topic?
Lopez: I've been working on data governance and ethics topics for a while, so it seemed logical to me to explore AI Ethics frameworks and guidelines. I've been collecting publicly available resources and doing comparing and contrasting them with each other. I decided I wanted to share what I found with others.

Can you give a quick example of how the EU's Ethics Guidelines for Trustworthy AI might be applied during an LLM development project?
That's an important topic I'll be covering. One thing that is critical to responsible AI is doing everything we can to mitigate bias in our training data, the models, and how we use those results.

For instance, most models have extensive training on data and resources available on the public internet, but a lot of the data in public areas aren't always the best quality for training because most complex, professionally developed examples are behind paywalls or internal firewalls.

The testing I've done on database designs shows that it can do basics well, but anything complex and enterprise-level suffers.

Do these frameworks provide any guidance on mitigating hallucinations in generative models, or is that still an open problem?
Yes, they do, often focused on how to do better prompting, especially around actually telling the system to only give verified information.

All of them do mention data quality as the first step, then including human verifications along the way, and finally educating users on how to avoid and spot hallucinations.

Is there a lightweight way to conduct an AI Ethics Impact Assessment without a large compliance team involved?
I believe the lightweight part of this question is might have different meanings to readers. There are assessment tools that can be used to get teams started quickly because they don't have to start from a blank page. They have checklists, templates, and other resources to help those of us who aren't auditors or legal experts to get started faster.

What resources can attendees use to learn more and prepare for your session?
The teams at Azure AI service blog have been writing about these topics for a while. The content used plain language to explain the concepts. I'd also recommend reading the public resources such as the EU, OECD, and Canadian government guidelines.

Note: Those wishing to attend the session can save money by registering early, according to the event's pricing page. "Save $400 when you register by the July 11 Super Early Bird deadline," said the organizer of the event, which is presented by the parent company of Visual Studio Magazine.