AI Runs on Public Data. That’s the Risk and the Responsibility.

As artificial intelligence (AI) has grown in use and popularity, my relationship with it as a working professional has already gone through several stages. The first was fear. As a journalist by trade and a communications consultant for more than 20 years, I worried I might have to reinvent myself late in my career. I’m sure there were and are many with similar fears across industries. Like many, I decided to get to know my “new colleague.” And the more I learned about its impressive capabilities, the more I found new meaning in what I do, and work in general.

While working with AI, it became clear that it is a collector. Every response comes from public information, including media coverage, executive commentary, websites, filings, and social content. When prompted to provide information on a company, AI acts as a mirror, assembling information from available records. For businesses, this means AI increasingly reflects an organization’s accumulated public footprint. When information is accessible, clear, and current, it presents a coherent picture; when it isn’t, gaps are filled, often with public partial truths and outdated content.

Today, AI summaries are influencing reputation, partnerships, talent decisions, partnerships, and even investment conversations. So how do we help AI capture our work accurately? And who, inside an organization, owns that responsibility when AI becomes an intermediary between companies and their audiences? The short answer is we all do. AI needs a Human-in-the-Loop (HITL), and this responsibility touches leadership, legal, HR, communications, operations, and finance, because AI’s “story” affects every part of the business.

AI is now embedded in workplaces across industries. As organizations rely on it to support decision-making, the quality of what it produces depends on the quality of what is publicly available. This makes accuracy, consistency, and credibility of public information a shared responsibility. Errors do occur, but they are rarely the result of AI inventing information. More often, they reflect fragmented, outdated, or inconsistent corporate narratives that have accumulated over time.

Take Athenahealth, for example, a healthtech company based in the United States. A recent New York Times article shared how Athenahealth discovered that AI chatbots were pulling outdated information and missing key details about their organization. Over six months, the company engaged in publishing targeted content, including updates to its website, social posts, and educational materials, to provide AI systems with accurate, current information to draw from. After publishing roughly 250,000 words of highly calibrated content, Athenahealth noticed that AI chatbots started pulling more reflective information of their company’s identity and work. While effective, not every organization can dedicate that level of effort, and smaller-scale, targeted updates can often achieve meaningful results.

In addition to keeping website content accurate and current and leveraging social media, we can influence the information AI systems draw by supporting journalism. Media, and the third-party validation it offers, plays a key and credible role in shaping how AI systems “understand” companies. Recent Muck Rack research shows that the vast majority (95%) of links cited by AI responses come from non-paid media, including news articles, with a significant portion published within the past year. Journalism doesn’t just inform audiences; it has now been inadvertently given a new role and responsibility of updating the source material that AI systems rely on to generate future answers. Journalism has become part of the data infrastructure that feeds AI-driven discovery.

And this role is becoming harder to sustain. As newsrooms shrink, illustrated by recent large-scale layoffs at the Washington Post and other outlets, the volume of rigorously reported, regularly updated source material that AI systems depend on is at risk of declining.

This creates new responsibilities and higher levels of accountability. Stewardship of accurate public information needs to be shared, now involving not just communications, but also legal, HR, compliance, finance, operations, and leadership. When organizations treat accurate public information this way, the public record improves, and the information that AI systems rely on is more likely to reflect an organization as it truly is.

In practice, this means prioritizing clear, consistent messages and aligning language across media coverage, websites, reports, social channels, and executive commentary so the public record is coherent for both people and machines. It means investing in initiatives that make accurate, credible information easy to find. These efforts help AI systems detect coherence rather than noise. They also include supporting high-quality journalism and credible third-party validation, which remain among the strongest signals AI relies on.

I was worried that AI would replace me. Instead, it clarified things. AI will report on a company whether or not a company is actively involved. This can have serious ramifications if we don’t collectively steward the truth. That’s why we all need to be the humans-in-the-loop, guiding and supporting credible information and journalism, so AI reflects data that is accurate and responsibly presented.

Julia M. Smith is a freelance writer, and the Managing Director of Finch Media, a public relations, strategic communications and health communications agency that provides flexible fractional support alongside full-scale communications solutions across sectors including automotive, education, entertainment, biotech, life sciences, and consumer packaged goods.