DEV Community

Bernát Kalló
Bernát Kalló

Posted on

Creating an AI policy

The gold rush has irreversibly begun, software companies are shoveling AI support into their apps, while experts in the field cannot scream loud enough about the risks. While there are infinite new opportunities, difficult to comprehend how immensely valuable, certainly there are also many dangers, which are maybe even more difficult to realize.

We as software developers and product managers have a HUGE say in what will happen and what not in cyberspace. We have way more power than we often realize. And because of that, we need to be very careful. And responsible. And we need to have principles.

Principles, by definition, should not depend on the actual opportunity. We should be prepared not to blindly follow the flow when a tempting opportunity comes if it compromises our principles. So I believe it's good to have an AI policy ready, preferably published on our website, so when I'm asked to create this or that AI integration, there is something written that I can refer to. A Ulysses pact.

So I've started to sketch an AI policy for my software development team, based on the bits that I've already understood about the risks of generative AI.

Here it is:

Our basic principle is that humans should rule over the computer and not the other way around. Therefore we do not create and do not install a system where

  • an AI makes decisions about humans, without a human taking responsibility about the decision
  • where an AI can uncontrolledly influence the outside world, or make uncontrolled actions on the internet

We think helping human editing work by AI is useful, however, we see the risks when the human editor only superficially verifyies the AI-generated content. Therefore, in the applications we develop, every longer (>~10 words) AI-generated content included into human-edited content will be marked with some subtle mark (color, comment). The human editor can remove this mark, thus explicitly taking responsibility for the content.

Similarly, when I code with an AI assistant, I strive not to accept longer (multi-line) AI-generated code completions at once, only line by line or word by word.

This is only my first attempt in this new genre, and I plan to further refine it as new trends are happening and my understanding of the topic broadens.

I encourage you to think about this difficult and sensitive topic. And please share your thoughts!

Image of Docusign

🛠️ Bring your solution into Docusign. Reach over 1.6M customers.

Docusign is now extensible. Overcome challenges with disconnected products and inaccessible data by bringing your solutions into Docusign and publishing to 1.6M customers in the App Center.

Learn more

Top comments (2)

Collapse
 
glszabolcs6 profile image
Gál Szabolcs #Luckylukee @GlSzabolcs6

Köszönöm a tartalmas összefoglalót, nagyon hasznos!

Collapse
 
cie profile image
Bernát Kalló

Köszi! :)

A Workflow Copilot. Tailored to You.

Pieces.app image

Our desktop app, with its intelligent copilot, streamlines coding by generating snippets, extracting code from screenshots, and accelerating problem-solving.

Read the docs

👋 Kindness is contagious

Immerse yourself in a wealth of knowledge with this piece, supported by the inclusive DEV Community—every developer, no matter where they are in their journey, is invited to contribute to our collective wisdom.

A simple “thank you” goes a long way—express your gratitude below in the comments!

Gathering insights enriches our journey on DEV and fortifies our community ties. Did you find this article valuable? Taking a moment to thank the author can have a significant impact.

Okay