Intro
As a design and tech studio focused on building a more equitable world, we are significantly limiting our AI use and taking an institutional stance against major AI companies.
- We will not use AI for generative creative work
- We are significantly limiting AI use in web & app development
- We will not be contributing financially to AI companies that are actively working to degrade democratic systems, have shown little or no concern for the environmental impact of their products, and whose leadership is putting forth a vision of the world with which we do not align
We’re a tech company, so we’re naturally interested in technology’s benefits. But one of our working values is “tech-skeptical,” which means making decisions that align with the best interests of humanity, rather than pursuing tech innovation for the sake of tech innovation. This is our first attempt at aligning our use of AI with that value. You can call our policy “AI Skeptical.”
This policy is going to strike some people as radical, or naïve.
Big Tech companies are investing millions in the narrative that full-scale adoption of their AI tools is inevitable. They want to dominate our economy. But the direction our economy takes is not inevitable; it’s still up to us.
Businesses like Outside influence the economy every day through our business model and use of resources. As a creative studio, the visuals and messaging we put out into the world shape perspectives and influence decisions. Even if Outside’s footprint is small, the collective power of all small independent businesses and creative studios is large.
Rather than succumbing to the narrative sold to us by Big Tech, we encourage businesses to form their own stance: What world would we like to participate in building?
If you believe in an economy fundamentally designed to prioritize the wellbeing and flourishing of all humans, and one that protects our environment and the resources required for life to thrive longterm, consider whether or not the tools you pay for are working toward or against that goal.
When more businesses have the courage to act on their values, we are more capable of forming independent ecosystems that operate on different rules with different outcomes.
Eventually, we create a new momentum. We determine how we use the tools tech companies create, not the other way around.
The direction our economy takes is not inevitable; it’s still up to us.
Outside’s Structure
Before explaining our policy, here’s a brief overview of how Outside structures our work:
Outside makes websites, apps, and brands. We are designers, web developers, and copywriters. We also act as a strategic partner to our clients, contributing to their messaging & communication strategies, helping roll out or plan new brands, naming products and programs, and more.
The work we do is essentially split between what we call ‘Creative’—visual design, branding, messaging, copywriting, and strategy—and what we call “Development,” or Dev, which is the actual building and maintenance of digital products such as websites & apps.
That distinction is important to understanding our AI policy.
All of our decisions as a business are guided by our values and ambitions. Read more about our Philosophy here.
AI Policy
Please note that this is a public-facing version of an internal policy. It is not exhaustive, and is expected to evolve as we continue to learn about the impacts of this tech.
Rationale for Unacceptable AI Use
- AI has an expanding environmental footprint, even if the per person footprint is small for now. In the U.S., AI has an embedded relationship with the fossil fuel industry.
- Most of the big AI companies are aligning themselves with political decisions and supporting politicians in a way that is antithetical to Outside’s values.
- Generally we try not to work with corporations that dominate industries, making our economy less democratic and regenerative. Big AI companies have massive valuations, and their leadership have monopolistic ambitions that make economies less democratic.
- AI acts as an accelerant of current capitalistic progressions, prioritizing profit, efficiency, scale, and productivity above all, which go against our goal of building regenerative economies.
- The creative output of AI is often generic and error-riddled. We think our Creative team produces more distinct, high-quality, hand-crafted words & visuals. Models are trained using peoples’ creative work—the product of years of practice, craft, and human imagination. AI is only regurgitating that work and effort into a generic outcome with no genuine intention, and we think that’s noticeable.
- AI use has been shown to lead to cognitive decay, diminishing our ability to think critically and impacting our personal relationships.
- Without significant human governance or guidance, AI amplifies biases embedded in data.
- AI companies use stolen data to train their models and then take work away from the people they stole from.
Cases where we will not use AI
We will not use AI to produce any creative work at Outside.
We value the humanity that powers and thrives from the creative process.
That means we will not use AI tools to create strategy, design, messaging, or copywriting. Everything Outside makes will continue to be made by humans.
We understand that there are instances where AI is already integrated into tools our Creative team uses—Google products, for instance. We don’t use Figma Make or Adobe Firefly, both AI-powered design tools connected to platforms we use.
We turn off any ‘Allow [platform] to train AI models from our design’ settings on our design tools. We ask members of our team to disable AI features when the features in question violate the acceptable use cases we outline here.
We will not maintain paid subscriptions to AI products made by companies that pursue de-regulation and personal financial gain above the best interests of the collective.
As of the writing of this policy, that includes not supporting those companies that have aligned themselves with Pres. Donald Trump’s administration.
Despite making claims about the potential good their technology may do for humanity, leadership at the most prominent AI companies have shown themselves to be singularly interested in financial gains and the pursuit of Artificial General Intelligence (AGI). To this end, they have allowed their tools to be used for acts of warfare, surveillance, and the distribution of disinformation.
Sam Altman (OpenAI), Elon Musk (xAI), Jensen Huang (Nvidia), Demis Hassabis (Google DeepMind), and others have pursued funding and their companies have reportedly lobbied for de-regulation, even when it meant political alignments that support war and genocide; the rollback of environmental protections; incursions on citizens’ privacy; and the disappearing and imprisonment of American citizens and immigrants.
Some key tools that we will not financially support include:
- Anthropic / Claude
- OpenAI / ChatGPT
- Google / Gemini
- xAI / Grok
These companies have shown themselves to be, in many cases, politically and values-agnostic; we think that’s dangerous.
Big Tech CEOs and corporations have reportedly donated millions to Pres. Trump, presumably in the hopes that doing so will mean less tech regulation and oversight. We do not think AI needs less regulation and oversight. We have already seen the dangerous ramifications of corporate influence in politics, both in the U.S. and Nepal.
Additionally, eight of the leading AI companies, including OpenAI, Google, and Microsoft, have signed agreements with this administration that will allow the U.S. military to use AI, which could lead to the development of fully autonomous weapons, as well as mass domestic surveillance that violates the rights and safety of all of us.
We will not use AI for tasks that would negatively impact a project’s performance or architecture.
We have seen instances where AI was used to speed up production and deliver short-term benefits, at the cost of a product, code, or project’s long-term performance, maintenance, and craftsmanship. We prioritize quality, and want to be proud of the work we produce.
Efficiency and perceived productivity thanks to AI does not justify losses in quality and / or performance.
As an example, AI may be able to generate code to produce a certain type of web animation. That said, the code produced may not meet our standards if it:
- Is not the most efficient way to achieve a particular animation, and therefore burdens the browser unnecessarily
- Does not achieve the optimum level of smoothness and refinement, and instead feels laggy or glitchy
Similarly, decisions around a project’s architecture and high-level feature logic may have serious implications on overall performance and maintainability of the stack long-term. Critical features and complex logic should be coded manually, as this ensures our devs are able to understand and maintain code longterm for projects we develop.
Rationale for Acceptable AI Use
- These are some instances where humans aren’t adding value to a process, and therefore there is no quality loss when using AI. These cases are minimal, but we have outlined them below.
- These are cases where the environmental footprint of using AI for a task is estimated to be the same or less than it would take to do the task manually, using a computer plugged into electricity. This is only possible when AI delivers high-quality results on first delivery, so the more complex the challenge, the less likely there will be energy efficiency through AI. Speaking broadly, generative AI has been shown to have a greater environmental impact than traditional AI use. For more about this, we recommend this article.
- These are cases where we aren’t providing the LLMs with crucial training data for further development.
Cases where we may use AI
We are currently researching both open-source and privately owned tools that align better with our values. We will update this policy once we settle into a consistent workflow with a specific tool. In the meantime, if we cannot achieve the acceptable use cases listed below using free tools, then we will not utilize AI in those cases.
We may use AI for assistance with code development and research, in certain instances.
Within a project, there may be more repetitive, non-critical, less complex elements that can be developed agentically, if prompted by mid-level and senior-level developers.
Junior developers should write all code manually, regardless of complexity, as this is crucial to their learning and understanding of the development practice. Due to their lesser experience, junior developers are more likely to require excessive and unnecessary prompting to complete a task without quality loss, which furthers the environmental footprint for any given task. Once these developers move beyond the junior level, at that point they may begin to also develop agentically for non-critical, non-complex tasks.
In any case, all AI written code must be manually checked by a developer for quality before moving to production.
All levels of developers, including junior developers, may use AI for assistance in coding research. For example, if we are trying to understand how to code a certain type of web animation, AI may be used to create an initial prototype to understand possible approaches, though the underlying code from the prototype should not be used to develop the actual animation in the given project.
We may use AI for non-creative or strategic administrative work.
Outside’s team members may use AI as a tool for a limited number of low-stakes manual, menial, or repetitive tasks.
For example, if a master spreadsheet needs to be broken into separate sheets corresponding to individual entries, we’ve found that an AI tool can achieve this task accurately and quickly, where it would’ve been time- and energy- consuming work to do manually. Outside team members may use AI for basic data-entry tasks like this.
We may use AI for background research.
AI is a massive data-computation process. One of the most basic, minimal impact uses of AI is as a tool to help sift through the vast amount of information available on the internet. We have used AI for this purpose, in the beginning stages of a project, to find specific articles, resources, and pieces of data that will improve our understanding of the problem our client is working to address.
When using AI for research, we require sources and we check them ourselves for accuracy. This is purely a background research task—we will not use any language generated by the AI tool for the final product of any of our projects. That would be considered “creative use” and would not be allowed in our policy.
What might change this policy?
We recognize that technology evolves rapidly, as does our understanding of its harms and benefits. Here are some conditions that may lead us to increase AI adoption in the future:
- A new LLM emerges, developed within an organization built on collectivist values with collectivist guardrails.
- Thoughtful AI regulation emerges, alongside actual enforcement of antitrust laws to check the power of AI / technology companies.
- We come to learn that some of our underlying assumptions about the harms of this technology and the companies that produce it were built on faulty / incomplete data or logic.
- The energy grid in the U.S. becomes significantly more renewable and no longer subject to Jevons Paradox, an economic theory about how an increase in a resource’s efficiency often leads to increased demand, offsetting any of the beneficial effects, such as cost-savings or sustainability.
