Why We Need to Move From Centralized AI to Federated AI

It’s no secret that large language models (LLM) owned by OpenAI, Google, and Meta are very expensive to build and maintain. After an estimated $4 million start-up to train AI models such as ChatGPT, we’re talking about at least $100,000 per month in maintenance.

Given the prohibitive costs, it makes sense for small and medium-sized businesses to piggyback off existing ChatGPT and Google models. Outsourcing AI innovation by relying on subscriptions offered by Google, OpenAI, or Amazon can help companies be more efficient.

From a business owner’s perspective, externally sourced AI saves money and enhances productivity. Who wouldn’t want that?

Here’s the catch… we’re helping Big Tech consolidate
According to the Small Business Association, 33.2 million small and medium-sized businesses (SMBs) account for over 99 percent of U.S. businesses. Yet, when it comes to AI, three companies dominate the U.S. market: Microsoft-backed OpenAI, Google, and Meta. That’s three companies currently serving 33.2 million customers and collecting their data with little to no competition. That’s also three companies growing bigger and stronger by the minute, thanks to SMB’s subscriptions and data contributions.

When you pay to use an LLM, you’re giving your data away for free
AI automatically gathers data from whatever online data sources are available using techniques such as web-scraping, web crawling, and APIs. Data is being grabbed from product descriptions, FAQs, blogs, social media posts, customer reviews, and pretty much anything that’s out there on the web.

When your company uses one of the big three LLMs, sure, you’re accessing a value-added service. But you’re also sharing your company’s and customers’ data with the LLM and furthering the power of Big Tech.

AIs are making it attractive for businesses to share their data. Once clients bite, LLMs use that data to create and control large data sets. From those data sets, they create new and improved AI models. The more data LLMs collect, the bigger, more accurate, and more robust models they can build.

Confidential and sensitive information is up for grabs
When you signed up to use AI to enhance your customers’ experience, did you intend to allow anyone to take or use your company’s data? Probably not. But, that’s what’s happening. Confidential information, client data, source code, and even regulated information are all there for the taking. There isn’t any visibility, accountability, or transparency on how the data is captured or leveraged. There isn’t even a way to request to delete information. If you have any trade secrets, those could be captured too.  

Large companies are learning from their mistakes
Recent leaks at Samsung are a prime example. Merely weeks after Samsung lifted a ban on ChatGPT, employees submitted source code and internal meeting notes to the chatbot. Since OpenAI saves all data points to improve AI models, the damage is irreversible. OpenAI now owns and can share Samsung’s confidential information.

Keep in mind that Samsung isn’t the only one sharing sensitive data. Health care, financial services, insurance, retail, advertising, hospitality, and companies across a broad range of industries are sharing their data too. A recent study indicated that 3.1 percent of customers surveyed who used AI shared confidential information.

You’re exposing your customers, too
Now, let’s address the elephant in the room. Privacy. Privacy issues go hand-in-hand with security breaches. There’s been a lot of noise about the recent ban on third-party cookies changing how companies request permission and collect information. Some European countries have recently begun investigating and banning AI’s use of private information. However, so far, AIs have escaped regulation in the United States.

ChatGPT makes no secret that they collect IP addresses, browser types and settings, and data about user interactions. They also collect information on what content users are engaging with, features used, and actions taken.  It’s all in OpenAI’s privacy policy.

Not to mention that deliberate attacks on AI systems are increasing.  Businesses like yours could end up spending millions to repair the damage. Let’s face it, no company wants to be put in that position.

Federated AI is closer than you think
Are we stuck with the status quo when it comes to AI? No. AI is still relatively new and evolving. Sure, a centralized world exists today, but at Fantix we believe that Federated AI is where businesses are heading.

Rather than centralizing data in a single location, federated AI techniques allow different organizations to pool their data while maintaining complete control over their resources. Multiple parties can train a model while maintaining data privacy. For instance, Fantix is working on a federated AI platform that will leverage our proprietary data abstraction technology. Anonymous representations of your data are extracted and encrypted, ensuring no raw data is ever shared.  

We believe businesses should be able to use AI without Big Tech watching
At Fantix, we believe that a federated AI paradigm is critical to ensure open, decentralized innovation, fair competition, privacy, and security of sensitive data. We’re biased, but when you look at the rise of groundbreaking privacy technologies such as data abstraction, multi-party computing, and federated learning, you’ll see that they are the keys to open access and privacy-safe data.

It's time to move towards a future where the concentration of AI innovation will not be in the hands of a few big tech players. The balance of power must shift towards a federated and decentralized ecosystem where thousands of organizations develop models without compromising privacy and security.