Guardians of AI: Adobe’s Grace Yee On How AI Leaders Are Keeping AI Safe, Ethical, Responsible, and True

As AI technology rapidly advances, ensuring its responsible development and deployment has become more critical than ever. How are today’s AI leaders addressing safety, fairness, and accountability in AI systems? What practices are they implementing to maintain transparency and align AI with human values? To address these questions, we had the pleasure of interviewing Grace Yee.

Grace Yee is the Senior Director of Ethical Innovation (AI Ethics and Accessibility) at Adobe, driving global, organization-wide work around ethics and developing processes, tools, trainings, and other resources to help ensure that Adobe’s industry-leading AI innovations continually evolve in line with Adobe’s core values and ethical principles. Grace advances Adobe’s commitment to building and using technology responsibly, centering ethics and inclusivity in all of the company’s work developing AI.

Thank you so much for your time! I know that you are a very busy person. Before we dive in, our readers would love to “get to know you” a bit better. Can you tell us a bit about your ‘backstory’ and how you got started?

I’ve been with Adobe for almost 20 years, taking on a variety of roles that have allowed me to grow and make an impact. Currently, I serve as the Senior Director of Ethical Innovation, where I lead global efforts to ensure ethics and accessibility are core to the innovations that we deliver to millions of customers around the world.

Before this, I led product development for Adobe Sensei, where I helped shape our AI strategy and made sure our platforms met the evolving needs of both creative and enterprise users. It’s been an incredible journey, and I’m proud of the impact we’ve made while staying true to our values.

You are a successful business leader. Which three character traits do you think were most instrumental to your success? Can you please share a story or example for each?

1. Ability to focus on actionable, practical outcomes: When we were developing Adobe’s AI Ethics principles, I was focused on ensuring they were concise, actionable and relevant. This focus has paid off — our principles have remained robust and effective for nearly six years, even as the AI landscape has evolved dramatically from generative AI to agentic AI.
2. A curious and forward-thinking mindset: I’ve always been fascinated by how things work and staying on top of cutting-edge innovation. That’s what led me to take on my current role — I wanted to be at the forefront of developing AI responsibly, and that decision has enabled me to build Adobe’s AI Ethics program from the ground up. My advice? Follow what intrigues you. When you’re passionate about your work, it shows, and when you focus on the future, you create opportunities to shape it.
3. Prioritize yourself: It sounds cliché, but I believe the most important person in your world is you. If you’re not happy, that energy impacts everyone around you. Only you truly know how you feel and what you need, which is why self-advocacy is essential at any point in your career.

Thank you for all that. Let’s now turn to the main focus of our discussion about how AI leaders are keeping AI safe and responsible. To begin, can you list three things that most excite you about the current state of the AI industry?

First, AI is revolutionizing creative workflows by making it easier than ever for people to bring their ideas to life. Generative AI tools are enabling artists, designers and marketers to focus on storytelling and big picture thinking rather than mundane tasks. It’s incredible to see how these tools, including Adobe Firefly, our family of creative generative AI models, are making creativity easier than ever before.

Second, AI is becoming more dynamic and user-friendly, thanks to advancements in responsive prompting. Models are getting better at understanding user intent, refining outputs in real-time and delivering results that align more closely with users’ needs.

Finally, agentic AI is an exciting new frontier, offering the potential to automate complex tasks and act as a proactive assistant — but proper oversight and safeguards are key.

Conversely, can you tell us three things that most concern you about the industry? What must be done to alleviate those concerns?

While AI is advancing rapidly, there are still critical challenges to address when it comes to responsible development and deployment. Most importantly, a lack of rigorous testing can lead to AI systems becoming unreliable or producing unintended harms. Companies must prioritize real-world testing with users to catch potential failures before deployment.

As agentic AI becomes more autonomous, ensuring proper human oversight is critical. AI should assist, not replace, human decision-making. Organizations must design systems that keep humans in the loop, including establishing a context-driven AI Ethics review process that aligns with their core values.

Finally, transparency in training data is crucial to building trust among users and mitigating bias. For example, with Adobe Firefly, we made a deliberate choice to train our generative AI models only on content Adobe has permission to use, giving our customers the confidence to create without concerns about copyright infringement. Addressing this concern will create AI that is innovative while also safe, fair and accountable.

As a leader leading an AI-driven organization, how do you embed ethical principles into your company’s overall vision and long-term strategy? What specific executive-level decisions have you made to ensure your company stays ahead in developing safe, transparent, and responsible AI technologies?

Adobe’s AI ethics principles — accountability, responsibility and transparency — serve as the foundation for how we develop and deploy AI. These principles are embedded into our governance process, ensuring that we proactively assess risks, engage different perspectives and maintain clear oversight of our AI’s real-world impact. To operationalize this, we created the AI Ethics Impact Assessment, a structured framework that guides teams in evaluating potential harms and establishing safeguards before launch.

We also made the decision to train Firefly only on content we have permission to use — underscoring our commitment to delivering commercially safe and IP-friendly generative AI that customers can use with confidence. By integrating AI ethics into our product development lifecycle and fostering continuous stakeholder engagement, we ensure that Adobe’s AI remains innovative, responsible and aligned with user needs.

Have you ever faced a challenging ethical dilemma related to AI development or deployment? How did you navigate the situation while balancing business goals and ethical responsibility?

When we first launched Firefly’s text-to-image model, we initially restricted the model from generating content based on the word “skeleton” as a precaution against potentially harmful or disturbing outputs. However, as users began engaging with the tool, we discovered this restriction was also unintentionally blocking harmless use cases — like creating skeleton images for Halloween. This presented an ethical dilemma: how do we protect against potential risks without unnecessarily limiting creativity?

To address this, we continue to refine our approach by introducing more nuanced content filters based on customer feedback that evaluate context rather than applying blanket restrictions. By staying engaged with our users, we were able to uphold our AI ethics principles while still enabling creative freedom.

Many people are worried about the potential for AI to harm humans. What must be done to ensure that AI stays safe?

Maintaining strong human oversight at every stage of development and deployment is important since you need to constantly test for safety. While AI can enhance efficiency and creativity, it should never operate without appropriate guardrails and accountability measures — this work is challenging, but hugely impactful.

At Adobe, we prioritize real-world impacts and outcomes because for us, it’s all about the human impact. We focus on high-risk, high-impact scenarios to ensure AI is safe and responsible. With agentic AI growing, we emphasize human oversight to prevent unintended consequences from increased autonomy. By taking a context-driven approach to ethics, we’re working to align AI with societal values, build trust, and minimize potential risks.

Despite huge advances, AIs still confidently hallucinate, giving incorrect answers. In addition, AIs will produce incorrect results if they are trained on untrue or biased information. What can be done to ensure that AI produces accurate and transparent results?

Ensuring AI produces accurate and transparent results requires rigorous testing, continuous feedback loops, and a strong commitment to ethical development. At Adobe, we’ve embedded these safeguards into every stage of Firefly development.

Our AI Ethics team conducts extensive bias and safety evaluations, while our dedicated IP moderation team proactively identifies and removes potential copyright risks. We also emphasize transparency by curating training data that Adobe has permission to use, including licensed content from Adobe Stock and public domain content where copyright has expired.

But AI governance doesn’t stop at launch; ongoing testing, real-world user feedback and industry collaboration are critical to refining AI models and addressing emerging risks. By taking this proactive and iterative approach, we can ensure that AI remains a reliable tool that enhances creativity while maintaining accountability and trust.

Here is the primary question of our discussion. Based on your experience and success, what are your “Five Things Needed to Keep AI Safe, Ethical, Responsible, and True”? Please share a story or an example for each.

1. Accountability: You need to have mechanisms in place to receive and respond to user feedback or concerns. For example, Firefly has undergone rigorous AI Ethics assessments to prevent outputs that could perpetuate stereotypes or fuel misinformation. This means fully understanding how AI is utilized in a feature and then identifying any potential risks associated with the feature. If we find or receive feedback on any concerns, we fix it.
2. Responsibility: You must have processes to ensure thoughtful evaluation and due diligence. At Adobe, we make it a point to integrate ethical considerations into any product development process and keep creator rights top of mind.
3. Transparency: It’s so important to be open with our customers about how we use AI and the data used to train our AI models. For instance, when we introduce new generative AI features, we provide clear explanations of how the AI is trained and ways we mitigate risks such as misuse or unintended biases.
4. Context: Context matters deeply when it comes to AI, especially in the creative industry. AI should always be developed with an understanding of its real-world applications and societal impacts. And it should be designed to enhance human creativity, not replace it. With Firefly, we provide context about the model’s capabilities and limitations, ensuring that users understand how to best use it while respecting the integrity of their creative process.
5. Scalability: Building a company-wide AI ethics governance structure is challenging, but companies must ensure they can scale effectively for the future. My team works closely across product, engineering, legal and other teams to integrate ethical considerations early in product development, ensuring potential harm and bias are mitigated from the start. In the first two years of using our structure, we evaluated over 100 AI features, and the next year, we hit that milestone in half the time. This growth reflects our ability to scale AI innovation without compromising our ethical standards.

Looking ahead, what changes do you hope to see in industry-wide AI governance over the next decade?

I hope AI governance will move beyond broad, one-size-fits-all frameworks and toward more context-driven approaches that account for how AI is used. A nuanced approach will enable companies to apply the right safeguards where they matter most, rather than imposing unnecessary restrictions on lower-risk innovations.

What do you think will be the biggest challenge for AI over the next decade, and how should the industry prepare?

The biggest challenge for AI over the next decade will be ensuring that AI technology evolves responsibly while keeping pace with rapid innovation. As AI systems become more autonomous, particularly with the rise of agentic AI capable of acting independently with minimal human supervision, maintaining human oversight will be critical. Without proper safeguards, these technologies could make decisions that have unintended and potentially harmful consequences.

You are a person of great influence. If you could inspire a movement that would bring the most good to the most people, what would that be? You never know what your idea can trigger. :-)

The industry needs to shift its focus from AI as a singular entity to evaluating its impact in specific applications — just as a resume evaluation tool carries different risks than an AI-powered font recommendation tool.

How can our readers follow your work online?

You can learn more about Adobe’s work on AI Ethics at our website.

Thank you so much for joining us. This was very inspirational.