Marketing Teams and the Unreleased Product Information That Ends Up in ChatGPT

Marketing Teams and the Unreleased Product Information That Ends Up in ChatGPT

Your marketing team is probably pasting unreleased product names, launch timelines, and pricing strategies into ChatGPT right now. Not maliciously. They are trying to write better press releases, brainstorm positioning, draft social copy, and build competitive comparisons. They are doing their jobs. And in the process, they are potentially destroying trade secret protection for some of your most sensitive commercial information.

CISOs tend to focus AI risk assessments on engineering and R&D, which makes sense given the volume of source code that has leaked into public LLMs. But marketing teams handle a different category of information that is arguably just as valuable and significantly less protected by technical controls: go-to-market strategy. Product roadmaps. Unreleased feature sets. Pricing models. Partnership announcements that have not been made public. Competitive intelligence. Launch dates tied to regulatory approvals or contractual obligations.

This stuff is trade secret material, and it is being treated like scratch paper.

Why This Qualifies as Trade Secret Information

Under the Defend Trade Secrets Act of 2016 (18 U.S.C. 1836), a trade secret is information that derives independent economic value from not being generally known and is subject to reasonable measures to keep it secret. The Uniform Trade Secrets Act, adopted in some form by 48 states, uses nearly identical language.

Unreleased product information checks both boxes easily. A competitor knowing your launch date, pricing tier structure, or feature roadmap six months early has obvious economic value. Courts have consistently recognized commercial and strategic business information as protectable. In Motorola, Inc. v. Fairchild Camera and Instrument Corp. (1973), the court held that marketing plans and pricing strategies constituted trade secrets. More recently, in Epic Systems Corp. v. Tata Consultancy Services (2016), a jury awarded $940 million (later reduced to $420 million) for misappropriation that included confidential business information beyond just source code.

The "reasonable measures" prong is where things get uncomfortable. If your company has no policy governing what employees can input into third-party AI tools, and your marketing team is routinely pasting product briefs into ChatGPT's default (non-enterprise) tier, you have a real problem arguing you took reasonable steps to maintain secrecy. OpenAI's standard terms of service for free and Plus tiers historically allowed use of inputs for model training, though they introduced an opt-out in April 2023. Many users never toggled that setting. And even with the opt-out, you are still transmitting confidential information to a third party's servers without a negotiated confidentiality agreement.

The Marketing Workflow Problem

Here is what actually happens in practice. A product marketing manager gets a confidential brief from the product team. It contains the working product name, target launch window (say, Q3 2025, timed to a specific industry conference), three pricing tiers with specific dollar amounts, a list of features including two that are patent-pending, and the names of integration partners who have not yet made public announcements.

The PMM opens ChatGPT and types something like: "Help me write a press release for [Product Name], launching September 15 at [Conference]. Three tiers: Starter at $299/month, Professional at $799/month, Enterprise at custom pricing. Key differentiators include [Feature A] and [Feature B], plus integrations with [Partner 1] and [Partner 2]."

Every piece of protected information in that brief just left the building. And nobody in security or legal saw it happen.

This is not a hypothetical. Samsung banned employee use of generative AI tools in May 2023 after engineers uploaded source code to ChatGPT on at least three separate occasions within a single month. Samsung's situation involved engineering, but the marketing version of this plays out constantly and with less visibility because marketing teams typically operate outside the technical monitoring infrastructure that security teams build around code repositories and development environments.

What Makes Marketing Uniquely Risky

• Volume and velocity. Marketing teams produce enormous amounts of written content under tight deadlines. AI tools are genuinely useful for this work, which means adoption is high and resistance to restrictions is strong.
• Aggregation risk. A single prompt from a marketer often combines multiple categories of sensitive information: product details, pricing, timing, partnerships, competitive positioning. Engineering prompts tend to be narrower in scope.
• Lack of classification habits. Engineers are generally accustomed to thinking about what is proprietary. Marketing professionals are trained to communicate information externally; the instinct is toward disclosure, not protection.
• Shadow IT prevalence. Marketing teams adopt tools quickly and often outside formal procurement. A 2023 Salesforce survey found that 55% of employees using generative AI at work had not received employer approval to do so.

The Legal Exposure

Losing trade secret status is not just a theoretical concern. It has concrete downstream effects.

If a competitor independently launches a similar product and you want to bring a misappropriation claim, opposing counsel will ask what measures you took to protect the information. Discovery of marketing team ChatGPT usage, with no governing policy and no technical controls, undermines your case before it starts. In Public Patents LLC v. MarchFirst, Inc. (2003), the court emphasized that trade secret holders must demonstrate "active, not passive" efforts to maintain secrecy.

There is also the partner and customer dimension. If your marketing team leaks a partner's unannounced integration, you may be in breach of your NDA with that partner. If pricing strategy tied to a government contract ends up in a public AI model's training data, you could face issues under the Procurement Integrity Act (41 U.S.C. 2102), which prohibits unauthorized disclosure of contractor bid and proposal information.

For companies in regulated industries, the exposure multiplies. A healthcare company's marketing team drafting materials about an unannounced product tied to a pending FDA submission is handling information that intersects trade secret law, securities regulations (if the company is public), and potentially HIPAA if patient data informs the marketing strategy.

What Reasonable Measures Actually Look Like

Courts evaluating "reasonable measures" under the DTSA and UTSA look at the totality of circumstances, but certain steps come up repeatedly in successful trade secret claims:

• Written policies that specifically address AI tool usage and identify categories of information that cannot be input into third-party systems.
• Technical controls that prevent or monitor transmission of sensitive data to unapproved AI platforms.
• Training that is role-specific. Generic "don't share confidential information" training is not sufficient when your marketing team does not recognize a pricing deck as a trade secret.
• Approved tooling that gives marketing teams AI capabilities within a controlled environment, reducing the incentive to use consumer-grade alternatives.
• Audit capability so you can demonstrate, if challenged, that you monitored and enforced your policies.

The goal is not to ban AI usage in marketing. That is impractical and counterproductive. The goal is to provide AI tools that do not require your team to send confidential information to third parties in order to get useful output.

How FirmAdapt Addresses This

FirmAdapt's architecture keeps data within your controlled environment. When a marketing team member uses FirmAdapt's AI tools to draft content or brainstorm positioning, the inputs do not get transmitted to third-party model providers for training or storage. This is a structural decision, not a policy toggle that individual users might forget to enable. It means your unreleased product information, pricing strategies, and launch timelines stay where they belong.

FirmAdapt also provides audit logging and role-based access controls that let compliance teams monitor how AI tools are being used across departments, including marketing. If you ever need to demonstrate in litigation that you took reasonable measures to protect trade secrets, having a record of controlled AI usage across your organization is significantly more persuasive than a policy PDF that nobody read.