Navigate the 2025 threat landscape with expert insights

Beyond the hype of AI lie its limitations

AI buzz is blazing through the cybersecurity space, but companies are realizing it may not be quite the game-changer that was promised. With the hope of increased efficiency, maximization of resources and lowered costs, AI still comes with notable limits within the weeds of detecting and responding to cyberthreats.

Consultant owner and doctoral candidate John Bambenek detailed "the promises and perils" of AI focusing on how the very nature of AI presents limitations. "AI is inherently retroactive. If there are good and bad changes, your statistical correlations will become inaccurate. AI cannot recognize what it has not been trained [on]." A good example of this is the start of the COVID-19 pandemic, where users' activity radically changed. Activity that was once malicious was benign and vice versa. "AI can't recognize what it lacks training data to recognize."

AI also runs the risk of overcomplicating an otherwise simple solution, particularly in situations like brute force: "If I'm getting 100,000 login attempts from the same IP address from the same username, I know I could probably block the IP address. So, if I could solve the problem with a firewall, why spend a bunch of money on AI?"

The questions surrounding AI still do not overshadow its enormous potential. Alex Holden, founder and CISO of Hold Security LLC, advises companies to regard AI with "cautious enthusiasm." "AI definitely makes life easier. But it does make a lot of mistakes, so we need to approach with caution and thought." Holden referenced examples of AI platforms sharing confidential information or providing wrong -- and potentially deadly -- instructions, such as when ChatGPT provided a garlic and olive oil marinade recipe for a user that turned out to grow botulism.

When the bad guys get their hands on AI

With the huge advancements made possible with AI, there's no question threat hackers want their piece of the prize. Rather than creating a new class of threats, Bambenek warned about the volume of threats increasing as hackers can use AI, like standard employees, to make themselves even more efficient. "The volume of attacks going up with AI is the real issue. A lot of attacks come down to arms races. If I send out a million phishing emails, 1% goes through 10,000 victims. If I create 1 billion emails across, more are getting through. More victims mean more money; more information is being lost."

Holden added that bad guys use AI as an attack framework in ransomware to figure out sensitive data -- the value of that sensitive data, who to contact, who to complain to turn this into payment.

"So, the bad guys are embracing AI for many different ways, and the way we are trying to make our lives easier, the bad guys [are] doing it for [this] purpose as well."

The new era of insider threats to watch for in 2025

Some of the most insidious company threats may lie right within company lines. Holden warned companies of the growing threat of hacktivists or disgruntled workers who disagree with some of the social and political stances their employees take and may act nefariously against them. In order to prevent hacktivism from striking, Holden advises companies to monitor social media activity on their employee devices and enforce social media code of conduct policies without hesitation.

Another stream of insider threat is the disgruntled worker motivated less by politics and more by money concerns or simply seeking revenge on the employers they feel have wronged them. Such actors can provide further information to malicious actors, heightening the need to monitor dark web activity.

Holden also warns companies of a much discussed but less addressed problem plaguing tech workers: burnout. Despite company advertisements about employee well-being and mental health, burnout remains a problem. According to this report from the National Alliance on Mental Illness, more than half (52%) of employees reported feeling burned out in the past year because of their job, and 37% reported feeling so overwhelmed it made it hard to do their job. These conditions certainly deter productivity, but when the burned-out workers are IT security staff, it can make a company more vulnerable to threats. "People unhappy and overworked may lead to an insider threat. People unhappy may lead to mistakes. People tired can click on dangerous attachments."

2025 is poised to deliver sticky cybersecurity challenges. But companies can face head-on the cybersecurity challenges in the coming year with realistic expectations and applications of AI, as well as remaining in tune and aware of their employees' activities and morale.

Alicia Landsberg is senior managing editor on the BrightTALK summits team. She previously worked on TechTarget's networking and security group and served as senior editor for product buyer's guides.