When you paste your draft email into an AI writing tool and hit rewrite, you've just shared that email — its content, tone, context, and potentially information about the people and business it references — with a third-party server. Most users don't think about this. Most privacy policies are written to minimise the chance that users will. This article explains what typically happens to your text when you use AI writing tools, what the actual risks are, and what architecture genuinely protects your data versus what's just marketing language.

What 'We Use Your Data to Improve Our Models' Actually Means. Many AI writing tools include a clause in their terms of service stating that user inputs may be used to train or improve their AI models. This is often buried several pages into the privacy policy with language like 'aggregate and anonymised data' or 'to improve our services'. What this means in practice: the specific text you type — your email drafts, customer messages, strategic plans, salary negotiations — can become part of the training dataset for future model versions. The 'anonymised' qualifier sounds reassuring but has limits. Researchers have repeatedly demonstrated that supposedly anonymised datasets can be de-anonymised when combined with other available data. The UK's National Cyber Security Centre explicitly advises organisations to consider what data is being shared with AI tools and what the provider's data retention and use policies are.

The Types of Information at Risk. The risk isn't abstract. Consider what a professional might put into an AI writing tool on a typical day: an email to a client about a delayed project (contains client names, timelines, internal context), a performance review draft (contains employee names, salary discussions, sensitive assessments), a proposal for a new business relationship (contains pricing strategy, competitive positioning), a message to a colleague about a challenging interpersonal situation, or a response to a legal or HR matter. Each of these contains information that, in any other context, would be treated as confidential. The fact that it passes through an AI tool doesn't change its sensitivity — but it does change who has access to it.

What to Look For in an AI Tool's Privacy Policy. When evaluating any AI writing assistant, ask these specific questions: 1. Where is my text processed — on your device or on an external server? 2. Is my text retained after processing, and for how long? 3. Is my text used for model training — this is often opt-out rather than opt-in. 4. Who has access to my retained data — employees, third-party processors, law enforcement? 5. What happens to my data if the company is acquired? The privacy policy that protects your data today may not survive a merger or acquisition.

Why Local-First Architecture Is the Only Real Answer. The cleanest solution to AI privacy concerns is architecture that doesn't require trust in a third party's data practices — processing that happens locally on your device with no server transmission. ForSocials is built on this principle. When you use the Rewriter, Chat with AI, or Templates features, your text is processed without being stored on any external server. Your Templates live in your browser's local storage. When you clear your browser cookies, they're gone — because there's no backup copy on a server somewhere. This approach has a trade-off: you can't access your templates from a different device, and clearing cookies means starting over. But for professionals working with sensitive information, this trade-off is clearly worth it.

The Specific Risk for Business Professionals. The GDPR creates specific obligations for businesses handling personal data — including the personal data of employees, clients, and prospects that might appear in AI tool inputs. If you're using a cloud-based AI writing tool for business communication, you may be transferring personal data to a third-party processor without the contractual protections that GDPR requires. This isn't a theoretical concern — regulators have been increasingly attentive to AI tool data practices as business adoption has grown. Local-first tools sidestep this problem by design: if data isn't transmitted, there's no transfer to regulate.

Practical Steps to Protect Your Data When Using AI Tools. If you're using AI tools that involve server-side processing, these practices reduce your risk: remove identifying information before submitting and replace names with placeholders; check whether your tool offers a business or enterprise plan with stronger data terms; use opt-out settings where available — most major AI tools have a model training opt-out that is off by default and buried in settings; and for the most sensitive communication, use local-first tools like ForSocials where your most sensitive drafts — HR matters, legal communications, strategic plans — are handled with AI assistance and no data exposure.

ForSocials stores nothing on our servers. Your text, your templates, your conversations — all local. See how it works.