*The above image was AI-generated.
In the grand tapestry of technological advancement, artificial intelligence (AI) emerges as both a marvel and a quandary. Its potential to revolutionize industries, enhance efficiency, and elevate human capabilities is undeniable.
Hold up a second – quick question. Could you tell that ChatGPT wrote the lede above?
If not, would you want to know? (Spoiler alert: It was).
That’s the central question facing organizations of all stripes who use generative and other AI systems—which at this point is almost every organization in some capacity—and who now understand the need to develop usage and disclosure policies around the use of the technology.
Generative AI is popping up in all the wrong places:
- Sports Illustrated saw its reputation tarnished when it failed to disclose that some of its writers were really AI.
- In another instance, a peer reviewed science journal published widely inaccurate images of rat physiology. In both cases, the reputational damage is clear.
- And the horror film “Late Night With the Devil” was recently roasted by some viewers after viewers noticed it had used AI-generated imagery without disclosure.
AI Disclosure, Labeling, and Usage Policies in PR
Some organizations in the PR and communications space have already released AI disclosure and labelling policies, including the Institute for Public Relations (IPR), which released its policy in early March.
IPR President and CEO Tina McCorkindale told Fullintel that the policy’s rollout was primarily about protecting the integrity of research, and was partially in response to seeing obvious AI-generated opening paragraphs in some research submissions similar to the one in this article.
“We were seeing instances of ‘AI speak’, where you can tell something is obviously written by AI,” she explains, adding that other obvious AI use cropped up in translated documents.
The IPR policy was partly inspired by an article back in January by PR professor Cayce Myers, McCorkindale explains, and articulates a set of best practices on how to use generative AI responsibly and transparently to preserve research integrity. It’s a living document that includes considerations to keep in mind when using gen AI, such as:
- Acknowledging that gen AI is prone to hallucinations (making things up).
- Acknowledging that AI models can be biased.
The policy also includes guidelines for use in IPR-published work, such as:
- A requirement that authors disclose how gen AI is used in published work.
- That gen AI is primarily used for editorial assistance.
It also includes rules around when the use of gen AI should be disclosed (and when it doesn’t need to be), and how to disclose, including:
- Gen AI should be disclosed when used in the research process, such as collecting or analyzing data or creating new content.
- Gen AI doesn’t need to be disclosed for topic generation, brainstorming, or minor editing.
- Disclosure should include the AI program, the prompt, and the section to which it was applied.
But it’s not just IPR that has recognized the need for an AI policy. The Public Relations Society of America (PRSA) released its own guidance for the ethical use of AI back in November of 2023.
In the government arena, New York City, Maryland, and Illinois all have laws on the books requiring AI disclosure when used in employee screening processes.
What Are the Dangers of Not Disclosing AI?
While proactive disclosure of AI builds trust with your audience and stakeholders, the dangers of uncontrolled and undisclosed generative AI are considerable:
- Misinformation, disinformation (either on purpose or driven by AI hallucinations), and narrative attacks are some of the most serious. IPR research shows that Americans are extremely wary of misinformation and disinformation. And there’s no doubt that the proliferation of large language models (LLMs) for generative AI will worsen the problem.
- Having proprietary company information used as training data for an AI model. If you were to input sensitive data to ChatGPT, for example, without explicitly telling the app not to use your data for training.
- Ethical and reputational risks associated with using generative AI and the public’s reaction.
- And then there’s the temptation of simply mailing in your next project and having a bot write the whole thing for you with little to no human supervision. That’s typically a recipe for disaster for many reasons we’ve mentioned above (and also pretty lazy).
“There’s an astronomical level of risk (around misinformation and disinformation),” says Chris Hackney, the founder and CEO of AI Guardian, an Atlanta-based company that offers AI governance and compliance software. “I think you’re going to see extreme situations of generative AI used at scale to persuade large numbers of people with manipulated content.”
The sheer scale and rapidity of potential narrative attacks against an organization, for example, means PR professionals will need real-time monitoring programs to keep on top of and mitigate these threats before they explode into trending topics.
What Should Companies Be Doing on AI Policy?
Some of Hackney’s advice for organizations that use generative or other forms of AI include:
1. Set up an AI committee. Companies should have a centralized AI center of excellence composed of a diverse cross-section of stakeholders to ensure centralized and transparent discussions around AI.
2. Develop a sensible AI policy: This means organizations should develop a policy that doesn’t just tell people what they shouldn’t do—the policy should also educate users on what they can do with generative and other forms of AI.
3. Map your AI projects: He says many CEOs have no visibility into AI projects or use cases at their organizations, which invites a level of risk for the organization. Ideally, this is done via a centralized software platform, not spreadsheets or email chains.
4. Establish AI approval criteria: Articulate the criteria by which your organization decides to proceed with an AI project (or not).
Conclusion: Generative AI Policies Are Here to Stay (For Now)
As AI programs (either used directly or embedded in other software) become more ubiquitous, it’s a near certainty that more organizations will develop similar AI use and disclosure policies. Doing so protects the organization and its employees.
However, whether that will always be the case is a matter for debate.
“At this stage in the game, people don’t expect everything to be written by AI,” explains McCorkindale. ‘But if there is a time where it becomes more ubiquitous and it just becomes part of our cultural norm, then maybe we won’t need these policies anymore because people just expect it. But we’re not (there yet).”
No doubt that’s where we are headed (or close to it). And if the intro to this article is any indication, it may be a while until we get there. But then again, maybe not: Check out how far AI-generated video has come in the past year.
