Prompt: a very secure space depicted as a lush but modern walled garden (Midjourney v6)

Security

Generative AI tools often use input data from users to improve their models. To maintain data security in our work, we ensure the AI tools we use are fully private from training. They live and stay within our ecosystems, allowing us to apply AI in our projects without the risk of exposing sensitive information.

Example

Most generative AI tools have a public base plan, but many offer enterprise versions with higher privacy and security.

ChatGPT Free plan allows user data to be used to train models.

ChatGPT Team plan guarantees user data is private from training.

Prompt: a daytime scene of a person standing in front of a concrete wall, a glass circular panel is in the foreground, partially obscuring and blurring the person (Midjourney v6)

Transparency

The rights to AI-generated outputs are complex. Currently, content created primarily by AI cannot be legally owned, and details used in prompting can easily infringe on intellectual property. We’ve found success using AI in early-stage concepting work, but we’ve also built mechanisms to flag infringement concerns. Throughout our process, we provide transparency in how AI is used and ensure our final outputs are free of ownership concerns.

Example

Prompts should be free of details, such as branded content, that may be infringing.

Prompt: Five Targaryen dragons weaving in a pendant like the ones in House of Dragon.

Prompt: Five snake-like dragon heads with glowing eyes, weaving around the border of the image.

Prompt: an overhead view of a modern city laid out in a pattern inspired by nature (Midjourney v6)

Accuracy

Generative AI is built on probability, predicting outputs based on training data. AI doesn’t truly understand content so it’s prone to hallucinations — false information presented with confidence. When applying AI in our work, we try to play to its strengths, avoid its weaknesses, and verify outputs with credible external sources.

Example

Prompts should be worded to leverage general knowledge, and specific factual responses should always be verified.

List the current top 10 US streaming services by number of subscribers.

What are some of the most popular streaming services in the US?

Prompt: an over-the-shoulder shot of a tech worker in their mid-20s using a modern mobile device (Midjourney v6)

Humanity

Generative AI is strongest as a partner, not a replacement, to human thoughtfulness, creativity, and problem solving. We bring constant evaluation on our use of AI to ensure it empowers people rather than diminishing them. We use it in ways that bring positive impact and avoid biases against individuals, communities, and groups. In our process, AI-generated content is a starting point, not a final solution.

Example

AI should be used collaboratively, not to replace a human role.

Use voice generation to pose as a live customer service representative.

Use voice generation to simulate a digital automotive assistant for voice interaction.

Responsible Tools

Responsible Tools


We carefully select third-party AI tools, and develop our own, to adhere to our values and guidelines, and those of our client partners.

We carefully select third-party AI tools, and develop our own, to adhere to our values and guidelines, and those of our client partners.

Spark

Our custom-made AI-powered Figma plugin, Spark utilizes technology from secure generative AI tools on the backend.

Get Spark

Secure & Cost-Effective

Spark’s secure version uses local and self-hosted Meta Llama3 models to ensure data is never shared.

Meta's Responsible AI Policy

Principles Built In

Spark tags generations with prompts and models, and flags prompts with potential for IP infringement.

Learn More About Spark

Third-Party Tools

Our teams utilize a primary set of third-party generative AI tools selected to align with our principles, but we’re constantly evaluating new AI tools and expanding our toolset while ensuring private and secure usage.

OpenAI ChatGPT & DALL-E

Enterprise ownership, control, and security of generations

OpenAI’s Privacy Policy

Midjourney

Stealth mode with generations hidden from the public

Midjourney’s Stealth Mode

Google Gemini

Enterprise-grade security and privacy within our workspace

Google's Privacy Policy

Adobe Firefly

Trained solely on licensed content with generations private from training

Adobe's AI Guidelines

Get

Smart

Interested in our generative AI tools and services? Let's work together to transform your vision into reality.

Let's Chat