Discussions on this topic surfaced repeatedly during the recent RSAC cybersecurity event, including a standout presentation, as well as at B-Sides 2026, an intimate gathering that occurred the prior weekend to RSAC. The reason centers on AI's vulnerability to deception, a weakness that malicious entities are actively exploiting.
Such interference doesn't solely involve grave threats, such as accessing data from personal computers. Rather, AI is often subtly directed toward actions that lack clear malice yet remain ethically dubious, according to Sherrod DeGrippo, who serves as Deputy Chief Information Security Officer and General Manager of Customer Security at Microsoft, in her RSAC 2026 remarks.
She illustrated this with buttons on websites labeled for AI summarization; upon activation, these covertly instruct the system to favor the site's products in subsequent suggestions. This differs from model poisoning, which involves injecting flawed training data. Here, the AI is simply prompted to follow undisclosed directives.
Risks encompass guiding users toward inferior or dubious items. Personal information might be gathered and traded to less reputable parties. However, the consequences tend to be indirect and gradual, primarily aimed at boosting sales. DeGrippo noted that this strategy has roots in past practices, where profit-driven individuals tampered with search algorithms to shape results, though its AI application is a recent development.
To counter these questionable methods, users should scrutinize AI-generated content closely, reviewing recommendations and overviews. DeGrippo observed that perpetrators of financial scams rarely innovate excessively; they apply just enough effort to meet their goals without further complication.
Given AI's swift evolution, unethical practices are likely to intensify alongside efforts to mitigate them. Staying informed remains essential for personal protection in this landscape.
With 15 years in technology and gaming media, Alaina Yee contributes to diverse subjects at PCWorld. A team member since 2016, her articles have spanned processors, operating systems, hardware assembly, browsers, single-board computers, and beyond, while she also scouts for deals (#slickdeals). Her current emphasis lies in cybersecurity, offering guidance on online safety. Past publications include PC Gamer, IGN, Maximum PC, and Official Xbox Magazine.