Let me be direct. The current system of unilateral Terms of Service modifications represents market power abuse dressed up as contract law. When a company can retroactively alter the conditions under which it accesses your data (data that may include your customer lists, proprietary processes, and strategic insights) we have moved beyond voluntary exchange into something approaching expropriation.
The free market operates on the principle that voluntary transactions benefit both parties. But when a provider can subsequently rewrite terms to grant itself rights to use your data for AI training, essentially converting your proprietary information into their competitive advantage, the original transaction has been fundamentally altered without your consent.
Yes, technically you "agreed" to terms that permit modification. But when switching costs make leaving impractical and critical clauses are buried in documents no one reads, that agreement is meaningless. This is a contract of adhesion, not a negotiation between equals.
This Is Not a Contract Dispute
It is an assault on property rights. Your data represents capital, accumulated through investment in customer relationships, research and development, and operational excellence. When a service provider claims the right to mine this capital for their own purposes, they are confiscating a portion of your business assets.
The problem is compounded by switching costs. Once a company has integrated its operations with a particular platform, the expense and disruption of migrating to alternatives creates economic captivity. Service providers understand this perfectly. They can afford to be cavalier about Terms of Service modifications because they know their clients cannot easily walk away. This is not competition. It is exploitation of market position.
You Are Subsidizing Your Competitor
Here is how it works.
Company A stores its customer data, operational processes, and strategic documents with a cloud provider. That provider subsequently uses information from its platform to train AI models. Company B, a competitor, purchases access to AI tools built on that same platform.
Company A's proprietary knowledge has been absorbed into models that competitors can now access.
This is not intelligence gathering in the traditional sense. AI models do not store and retrieve your specific documents on demand. But patterns from millions of inputs (including yours) inform how the model responds to similar queries from anyone, including competitors in your industry asking questions about problems you have already solved.
The aggregate effect is the same: value flows from those who create proprietary knowledge to those who simply subscribe to the platform that absorbed it.
This Is Not a Collection of Isolated Incidents
It is an industry-wide shift. Over the past two years, major platforms have quietly rewritten their terms to claim rights over user data for AI training.
X (formerly Twitter) updated its terms of service in October 2024, effective November 15th. The new terms grant X a worldwide, nonexclusive license to analyze and use all content shared on the platform for AI training, including personal posts and any content you share. Users who previously had opt-out options found those protections unclear or eliminated in the new terms.
LinkedIn did not even update its privacy policy before it started scraping user data for AI training. When reporters confronted them in September 2024, LinkedIn said it would update its terms "shortly" and then quietly released a blog post the same day announcing changes. The company admitted it had been using users' data to train its AI without consent. Users can opt out of future training, but there is no way to undo past data use.
Meta announced in June 2024 that it would use public posts from Facebook and Instagram to train Meta AI. In the US, Meta had already been using public posts for this purpose. Users in the US have never been given the opportunity to opt out of having their public posts used to train the company's AI models.
Google updated its privacy policy in July 2023 to explicitly include AI training. The policy now states that Google uses publicly available information to help train its AI models and develop products like Google Translate, Bard (now Gemini), and Cloud AI capabilities. Any content publicly uploaded online could be used in training current and future AI systems.
What began as services for hosting and sharing content has evolved into data extraction operations for AI development. The terms just took a while to catch up. In some cases, the extraction started before the terms changed at all.
The Legal Landscape Offers Little Comfort
Courts have recently ruled that AI companies can use copyrighted books to train their models under "fair use" doctrine, deeming the practice "exceedingly transformative." At the same time, Anthropic agreed to pay .5 billion to settle claims over pirated training data.
The legal framework is contradictory and evolving. Waiting for regulators to protect you is not a strategy.
Act While You Still Can
Demand contractual protections upfront. Insist on clauses that prevent unilateral modification of data usage rights. If a provider refuses, find one that will. Enterprise agreements often offer stronger protections than consumer terms, but only if you negotiate them before signing.
Treat vendor diversification as a business continuity strategy, not an IT preference. Just as you would not put all your financial assets in a single bank, do not entrust all your data to a single provider.
Recognize that data storage agreements are ongoing relationships requiring active management. The cost of a service includes not just the monthly fee but the ongoing risk of terms modification. Monitor those terms. Assign someone to read the updates.
Audit what you are already exposed to. Your employees are using tools right now. Do you know which ones? Do you know what those tools' terms say about data usage? If the answer is no, you have a problem that already exists, not a future risk to plan for.
The Choice Is Yours (For Now)
Make no mistake. This issue will determine whether the AI economy operates according to principles of property rights and voluntary exchange, or becomes a system where a few large platforms systematically harvest value from the businesses that depend on them.
The window for action is closing. Every month, more platforms update their terms. Every month, more of your data gets absorbed into models you do not control.
The choice is yours, but only if you act while you still have the power to choose.
Raphael Savastano is the founder and principal consultant of ROFONIC LLC. With 25+ years in IT, 16 years in leadership, including 8 years as CIO scaling technology for a global manufacturer from M to 0M. He now helps growing companies get executive-level technology and operations leadership without the full-time cost. Want to know where your technology actually stands? Take the Founder's IT Reality Check →
