Early versions of AI were impressive, but completely disconnected from tools and data. GPT-3.5 couldn’t consistently add 2+2 or report on current events, but wrote competent literary criticism based on its extensive training on academic writing. AI was fascinating, but only had a modest impact.

But since then, the industry has been working overtime to bridge the worlds of AI and data. Today, companies large and small deploy agents throughout their operations. These agents can access enterprise data to help people make decisions and, increasingly, themselves take direct action on behalf of users. For example, when your flight is canceled, an airline’s AI agent can find you another flight, rebook your ticket, and send you a new boarding pass, all without you having to call customer service.

Agents are getting deployed to every one of us as consumers, and they will increasingly gain access to highly sensitive personal data. I recently connected my watch's health metrics and my financial data to my personal agent. A weird experience, but AI is the best personal trainer and financial planner I’ve ever had. Billions more humans will do the same.

[CTA_MODULE]

Stewarding data in an AI world

Amid heated public conversations about safe and ethical AI use, we don’t want to lose sight of the fact that data unlocks AI’s superpowers.

We believe AI will continue to progress, reshaping infrastructure and jobs, and alignment will be hard but achievable. Safely navigating these challenges is incredibly important and will take sustained effort and attention from the public, industry, and policymakers.

But none of this changes the fact that data forms the foundation for all this progress. The entire business of Fivetran + dbt Labs is to connect data with AI, and we think that there are real right and wrong ways to go about doing this.

As policymakers and industry leaders debate what comes next, we propose the following 5 public policy principles for data and AI:

People come first: People should know how their data is used and be able to access or delete it, and everyone who handles it should protect it.
Your data, your choice: Data should stay in the control of those who create and rely on it, not the systems that store it. Switching providers shouldn’t mean losing your data or paying unreasonable fees to get it back.
Open by default: Data should be portable, not trapped in closed formats or systems. Open standards keep markets competitive and give new ideas a real chance.
Trusted data is the foundation: Reliable, well-governed data is what makes AI and every other technology accountable.
Rules that work for everyone: Privacy and data rules should protect people everywhere without creating costs only the largest companies can afford. Good rules can be sophisticated without being complicated, so a 5-person startup can follow them as easily as a global enterprise.

We’ll keep building on these principles as the conversation evolves.

While public attention is on what AI can do, we want to make sure there’s just as much attention on the data foundation that makes it possible.

Because getting AI right means getting data right.

[CTA_MODULE]

Fivetran + dbt Labs is launching a new public policy initiative.