AI biases, benefits and what lies ahead for data privacy in Maryland

As emerging tech systems ingest more data and state-level regulation takes effect, safety is more important to businesses than ever.

Data privacy isn’t just a technological issue — it’s a business, legal and ethical challenge that affects everyone.

With new privacy regulations, the Maryland Online Data Privacy Act (MODPA) on the horizon and the speed with which AI evolves, this year’s Data Privacy Day on Jan. 28 felt timely. Taking place within a few industry- and policy-shaking developments — including DeepSeek’s announcement that it could outperform ChatGPT prompting a massive single-day selloff, the Trump administration freezing and unfreezing federal grants, and Alibaba announcing its own AI model — an event last Tuesday highlighted the voices of local experts while exploring data privacy’s prospects for 2025.

The gathering was held at UpSurge Baltimore’s offices in the Mid-Town Belvedere neighborhood, and organized in collaboration with Ardent Privacy, the International Association of Privacy Professionals, bwtech@UMBC, the Cybersecurity Association (previously Cybersecurity Association of Maryland, Inc.) and the University of Baltimore. Here are a couple of takeaways for anyone concerned about how data privacy and artificial intelligence will play out this year.

We’ve been through something like this before

CEO Sameer Ahirrao of Ardent Privacy, a company based at bwtech@UMBC, compared today’s pattern of AI innovation to the data-breach-ridden era of early internet adoption in the 1990s and 2000s. Back then, regulations tightened only after repeated, high-profile hacks prompted widespread concern. He predicted that AI data privacy would follow a similar pattern.

Sameer Ahirrao further noted that although tools like ChatGPT boost productivity for many businesses, the gains come with large risks, consumer trust challenges and increasingly strict regulations on data collection and usage.

Rather than expending energy on changing the behaviors of the enterprises behind AI applications like ChatGPT and Microsoft’s Copilot, Ahirrao recommended that businesses instead focus efforts on identifying what data flows to those models and creating an appropriate governance system (which is what Ardent Privacy does).

MODPA is strong, but uncertainties and social justice concerns remain

Sameer Ahirrao participated on a panel that explored “Artificial Intelligence or Artificial Risk,” how Maryland’s upcoming privacy law might shape data governance and to what companies can look forward.

Moderated by Joel Benge, a consultancy founder and Techstars mentor, the session also featured University of Baltimore law professor Michele Gilman; Laura Mateczun, the interim associate director of the Maryland Institute for Innovative Computing at the University of Maryland, Baltimore County; and Maureen Dry-Wasson, global privacy officer at the Maryland staffing agency Allegis Group.

Gilman described how MODPA could be one of the most comprehensive data privacy laws in the nation. She emphasized tenets like data minimization, which she said requires “ensuring that companies collect data that are reasonably proportional to carry out the task” for which customers engage them, and broad protections of sensitive information. However, only the Attorney General can enforce the law, and there are concerns about resourcing and prioritization.

While strong on paper, the impact will hinge on how effectively it is enforced when MODPA takes effect in October, even though enforcement won’t start until April 2026.

Gilman also stressed that the most impacted groups, usually low-income and marginalized communities, are often not part of the design process. She encouraged companies to incorporate diverse perspectives early to prevent AI systems from augmenting existing inequalities.

AI cannot be perfect

Dry-Wasson acknowledged the concerns around AI data privacy, governance and potential harms, but returned to the technology’s benefits. The Allegis Group uses AI to sort through millions of resumes and match “the right people to the right jobs at the right time,” she said. Using AI speeds up the review and matching process, but can also perpetuate biases if not carefully managed.

“We cannot compare AI to perfection,” Dry-Wasson said. “People are innately biased. When you have people doing things, there are just as many problems as with AI. The problem with AI is that it scales, it scales quickly and largely. So we have to be very careful about it, but we also need to remember we’re not comparing it to perfection.”

Gilman responded by saying that AI-driven processes can make true responsibility harder to enforce.

“There is a difference between human decision-makers and AI — not necessarily better or worse, as you rightly pointed out, but accountability is much easier in an analog world than a digital, black box world without transparency,” she said.

What can people do next with this information?

Speakers offered several strategies for local leaders, entrepreneurs and technologists to comply with upcoming regulatory changes:

• Audit your data. Identify the data collected, why it’s needed and where it flows. This helps teams decide on appropriate tools and security measures.
• Incorporate “privacy by design” early on. It’s far better (and more affordable) to start this way, rather than retroactively updating and outfitting solutions into a patchwork solution later on.
• Prioritize inclusive design. Input from broader and more diverse perspectives reduces blindspots, biases and harms.
• Prepare for slapdash oversight. States are enacting their own laws because the federal government has not yet created comprehensive and standardized privacy regulations. This will likely continue, and businesses must stay agile to adapt quickly.
• “Transparency … and documentation, documentation, documentation,” said Mateczun. AI systems already shape operations, research and decisions; documentation of these instances is necessary for understanding and improvement. That’s what MODPA is for: It’s “designed to push us towards documentation,” she said.