Over the recent month, a string of cybersecurity incidents involving AI has occurred. All facets of AI were touched, from releasing confidential training sets via cross-poisoning private models by incomplete permissions to leaking secret keys by code completion tools. None of these even graze the mirage of enhanced attacks, made possible thanks to AI attack tools.
Consequently, the first analysts call for a unified AI and cybersecurity strategy. While strengthening cybersecurity is always a good cause, whether a combined approach is a good idea remains questionable. Thus, let us dive deeply into the recent incidents and whether making them part of a unified IT Strategy is wise.
The String of Cybersecurity Incidents: Uptick or Reporting
Recent news has shown several incidents involving AI, from malware built using ChatGPT to optimizing password theft through social engineering. However, with many incidents, such as the data breaches mentioned in the introduction, the problems are within the AI space but don’t involve AI. Yet, the increased reporting and interest in AI cause these painfully common incidents to be picked up and distributed widely. It thus remains with the experts to differentiate between genuine AI threats, those attacks that get supercharged, and issues within its vicinity.
Does AI change cybersecurity?
This analysis, however, should not dismiss AI’s impact on cybersecurity. Apart from emerging and enhanced threats, the technology also changes the equations for defense.
AI products and technology can enhance detection and optimize machine learning. The changes are especially noticeable at the interface between humans and computers, such as in spam detection.
AI also does a great job in automation. Most of us follow routines, and AI is great at detecting changes from the practices. Thus, if the log suddenly shows unusual logins, it can alert staff or block them outright. The difference is that abnormal patterns refer to each employee’s behavior and not to some grand, all-encompassing policy.
Thus, AI will change both the offense and the defense and challenge us in new ways.
Cybersecurity separate from AI Strategy
Given the changes to cyberattacks and cybersecurity, the call to unify the two seems reasonable at first. However, many defenses, from training to gamification, do not involve AI. On the other hand, digital transformation, especially in automation and entry-level customer service, can significantly benefit from AI. Likewise, unifying the two might restrict the talent pipeline when searching for experts to work on cybersecurity and AI.
While we cannot escape cybersecurity, and all businesses need to consider it at a certain level, the same is not true about AI. A plumber must protect his address book, scheduling, and E-Mail. Yet, he has little need for AI.
Even for AI businesses, there are areas that AI doesn’t touch, but cybersecurity does.
Take time for each, instead of losing in both
AI is enhancing cybersecurity attack vectors. However, it is not an all-encompassing solution to fixing cybersecurity. With most attacks still relying on human factors, believing that any technology will ever solve it is a fallacy.
While management and boards must allocate time, budget, and expertise to Cybersecurity and AI, they are not interchangeable. Having the time for both might be tough, yet it shouldn’t be an excuse to lose out on both.