How to Use Generative AI for Content Without Legal or Security Risk

To use generative AI for content creation without exposing your company to legal or security risks, enterprise marketing leaders should evaluate AI solutions across four dimensions: data security, customer-centricity, accountability, and scalability. The safest path pairs human subject matter experts with AI — using generative tools to adapt and scale content originated by humans rather than generating content from scratch — while choosing enterprise-grade tools that prevent proprietary data from being stored in or exposed to public AI training models.

Subscribe to our monthly newsletter, Content & Context, to receive more insights on the evolution of content from Skyword CEO, Andrew C. Wheeler.

According to a recent Gartner poll of over 2,500 executives, 70% are actively investigating and exploring generative AI, while 19% are already in pilot or production mode. This paints a vivid picture of the business urgency that's built up around generative AI this year alone.

As with the eruption of any new technology, the critical challenge for business leaders is balancing initially inflated expectations with the realities of real-world application.

While 68% of the execs polled by Gartner believe the benefits of generative AI outweigh the risks, it's forecasted that trust, security, privacy, and ethical challenges loom on the horizon for many organizations developing and deploying gen-AI solutions.

The most anxiety-ridden question I encounter among marketers is how to safely and responsibly leverage generative AI for content creation in light of tech uncertainties. 

The risks are documented and growing

• OpenAI's ChatGPT, Microsoft's Bing, and Google's Bard are all proven to generate false information. 
• Generative AI tools, including image generators, have a well-documented tendency to generate discriminatory outputs.
• Foundational models are often trained on copyrighted works but don't provide reliable sourcing in their results.

But how real are these risks for marketing organizations, and what can you do to mitigate them? Here's my advice to those of you weighing your options: 

Three paths to generative AI adoption for enterprise marketing

1. Public AI Tools: These tools, including Bard and ChatGPT, are readily available to the general population. They are cloud-based platforms that allow users to interface with pre-trained AI models directly in an unstructured environment. They're low-cost and easily accessible but also the most prone to risk and misuse. Going this route means it's up to your business to develop and maintain the surrounding systems necessary to ensure these tools are securely, responsibly, and consistently adopted into your workflow.  
2. Proprietary AI Tools: These tools are developed for and owned by your company. Building vs. buying can be a good choice if you have unique requirements or need total control over the AI models' training data and development. However, the significant costs, effort, and ongoing data required to effectively train and maintain AI models in this way mean this option is out of reach for many organizations. 
3. Enterprise AI Tools: Designed for large-scale organizations with complex marketing needs, these tools provide structured AI applications, scalability, integration with existing systems, and baked-in security features.

The caveat: We're experiencing an arms race among tech companies to claim dominance in the AI sector. OpenAI, for example, just announced its forthcoming ChatGPT Enterprise, which has been under development for less than a year. Exercise caution with vendor selection to safeguard your organization from added risk and uncertainty, as many new "enterprise" AI tools are hastily entering the market without oversight or regulation.

Four criteria for selecting a generative AI content solution

Whatever option you pursue, crafting a responsible approach to AI-powered content creation is essential. In this vein, lots of articles explain what not to do, so I thought I'd share my four DOs for marketers instead:

1. Protect your data: take security seriously

Proprietary and enterprise tools tend to offer the most secure and controlled environments for tapping into generative AI models. With advanced control over user access and data privacy, these options will help keep sensitive data contained within your organization's infrastructure and safeguard you from any unlawful or unwanted third-party exposure.

Skyword's AI-powered content engine Accelerator360™, for example, ensures your proprietary information is never stored, exposed, or used in AI training models. By contrast, OpenAI's ChatGPT interface saves input data by default, including chat history, and the company may use the information to train and improve its models.

Ensure your solution(s) address:

• How user access will be securely managed and unauthorized access prevented
• How data inputs and IP are stored and handled, including for algorithm training
• What safeguards are in place to protect your brand from misuse and external attacks

2. Prioritize your customer: don't let AI flood the market with noise

For most companies, the need to rapidly scale content creation is the biggest driver of generative AI adoption. However, achieving this is entirely counterproductive if it comes at the cost of your customers' experience.

In 2013, Velocity Partners published a viral Slideshare titled Crap Content. It's embedded below if you want to page through, but the TL;DR is this:

When content marketing took off, everyone jumped to create as much content as possible to stay competitive. Except...they didn't. 

Because they didn't have the bandwidth or the appropriate resources to create quality content, all they did was saturate the market with noise that overwhelmed buyers and did the exact opposite of what it was intended to do.

Image Attribution: Crap. The Content Marketing Deluge. From Velocity Partners.

The companies that excelled at that time built content brands that provided visitors and buyers with value-added content and information. The same holds true today.

The mainstream availability of public generative AI tools threatens to usher in a new era of crap content. Meanwhile, customers' expectations for accessible, helpful, and relevant content are even higher. 

The aim of your solutions approach shouldn't be to create more content. It should be to create content of superior quality or greater strategic value than what you're producing today. For now, the most reliable way to create high-quality, original, and authentic content with generative AI is to leverage it in tandem with human creators and/or subject matter experts who can properly steer it. 

In previous posts, we've outlined many ways our generative AI solution and others can add unique value to human creation.

Ensure your solution(s) address:

• Long-term viability. Is your solution an operational band-aid, or does it grant you a competitive edge as content competition and customer needs evolve?
• The origin and originality of content. Remember, your content isn't copyrightable unless it's solely or primarily generated by humans.
• The need for audience and brand customization, ensuring your content is differentiated from a potential wealth of similar, AI-generated information being published

3. Build in accountability: know who owns the output

If you're building a proprietary system, it's on you to maintain comprehensive documentation of the AI model's development, training process, and deployment, providing insights into how the model was created, how data is used, and its decision-making logic.

Meanwhile, creating content using external tools requires careful contingency planning. Be curious about how the sausage is made and what strategies are in place for handling situations where the AI system inadvertently produces incorrect or potentially harmful outcomes.

Consistent human oversight is critical to effectively using generative AI, but beware that many tech vendors will hand over an enterprise license and leave this kind of oversight to you.

The last thing you want is to deploy a solution to save your marketers time on the front end only to have them sucked into a black hole of editing and fact-checking on the back end. 

For this reason, our solution leverages generative AI to adapt content originated by human experts before sending it to our in-house editorial team to review for quality, accuracy, and credibility.

A solid enterprise partner will help your team develop content while managing data governance and, ideally, offering continuous monitoring and auditing, model introspection, and human oversight and intervention — saving you valuable time and dramatically reducing the risk of error. 

Ensure your solution(s) address:

• How quality assurance/human oversight will be managed
• Who will be responsible for brand compliance
• Who will be responsible for the reliability of content inputs and outputs
• How ongoing optimization will be handled as the technology advances

4. Validate scalability: make sure your approach can grow with you

What's scalable for one organization may not be for another. The hard and fast rule here is to make sure your solutions approach is flexible and sustainable enough to evolve with your organization for at least the mid-term. 

Are you introducing a stop-gap process or procedure to be leveraged in a discrete corner of your organization, or are you putting a building block in place — a tool or tech-enabled process that sets you on a path toward enterprise transformation?

The allure of enhanced creativity and productivity through generative AI is compelling. But there is a delicate balance between chasing incremental productivity gains and leveraging AI to actually enhance your content's value, achieve meaningful reach, and secure ongoing increases in engagement.

Ensure your solution(s) address:

• The potential to sustainably expand usage across a broader number of teams and use cases over time
• The support mechanisms in place to ensure consistent, widespread organizational adoption
• The foundational expertise of your solutions partner(s) with respect to your highest-priority use cases 
• The level of commitment your in-house developers or external partner has to evolve their generative AI capabilities

The next frontier for AI-driven content marketing

It's important to recognize that we're all on a path to AI-driven transformation, whether we care to accept it or not. 

What will ultimately separate the leaders from the laggers is owning up to the responsibility to prepare our teams and organizations for a future of marketing in the post-AI era. 

It's never been a more exciting time to be in marketing. Let's savor the moment and step confidently into the future.

Is generative AI a priority for your organization this year? Drop me a line. I'm here to help.

Key Takeaways

• Enterprise AI tools with baked-in security are the safest path for content creation at scale — they prevent proprietary data from being stored in or exposed to public AI training models, unlike open tools like ChatGPT that save input data by default.
• The most reliable way to mitigate legal risk is to pair generative AI with human subject matter experts — AI adapts and scales content originated by humans, keeping it copyrightable, accurate, and brand-differentiated.
• Vendor selection requires caution: many "enterprise" AI tools are entering the market hastily without oversight or regulation. Evaluate providers on data governance, quality assurance processes, and ongoing monitoring commitments.
• Accountability must be assigned before deployment — determine who owns quality assurance, brand compliance, and content reliability to avoid transferring your team's time savings into a back-end editing and fact-checking burden.
• Scalability is not just about volume — the goal should be content of superior quality and strategic value, not more content. Solutions that merely automate output risk recreating the "crap content" problem that overwhelmed buyers in the early days of content marketing.

Frequently Asked Questions

Q: What is the biggest security risk of using public AI tools like ChatGPT for enterprise content creation?

A: Public AI tools like ChatGPT save input data by default, including chat history, and the provider may use that information to train and improve its models. For enterprise marketers, this means proprietary information — brand strategy, competitive intelligence, unreleased product details — can be stored, exposed, or ingested into a public training model. Enterprise-grade tools with secure APIs prevent this exposure.

Q: Can AI-generated content be copyrighted?

A: Under current guidance, content generated solely or primarily by AI cannot be copyrighted or owned by your brand. To maintain copyright protection, the most reliable approach is to have human subject matter experts originate the content and use generative AI to adapt, scale, or repurpose it for different audiences and channels.

Q: How do I evaluate whether an enterprise AI vendor is trustworthy?

A: Evaluate vendors on three dimensions: how they store and handle your data inputs and IP (including whether data is used for algorithm training); what human oversight and quality assurance processes are built into the workflow; and their commitment to ongoing monitoring, auditing, and model improvement. Be especially cautious with new entrants — many "enterprise" AI tools have been in development for less than a year.

Q: Should we use generative AI to create more content or better content?

A: Better content. The mainstream availability of generative AI threatens to flood the market with low-quality material, much as the early content marketing boom did. Customers' expectations for accessible, helpful, and relevant content are higher than ever. The aim should be to create content of superior quality and strategic value — not simply more of it.

Featured image by master1305 at Adobe Stock.