Artificial Intelligence (AI) has shown vulnerability to manipulation, raising concerns within the cybersecurity community. This was a significant topic of discussion at the recent RSAC cybersecurity conference and the preceding B-Sides 2026 event. The ability to trick AI systems is being exploited by various actors, not always with malicious intent, but often to achieve questionable outcomes.

As Sherrod DeGrippo, the Deputy Chief Information Security Officer at Microsoft, highlighted during her presentation at RSAC 2026, the manipulation of AI can occur through seemingly harmless tactics. A prime example she presented involved website buttons labeled Summarize with AI. Upon clicking these buttons, the AI is covertly instructed to prioritize specific products from the brand, thereby skewing future recommendations in favor of those items. This practice does not constitute poisoning the model with harmful data but rather covertly directing the AI’s responses without the user's knowledge.

The implications of such manipulation can be significant. Consumers may inadvertently be guided towards inferior or dubious products, and their personal data could potentially be harvested and sold to less scrupulous buyers. While the immediate risks may not be overtly harmful, this type of manipulation often focuses on increasing commercial revenue. DeGrippo noted that such tactics are not new; similar schemes have long been utilized to game search engines and influence online recommendations.

To avoid falling victim to these dubious practices, it’s crucial for users to remain vigilant regarding the output generated by AI systems. Scrutinizing the suggestions and summaries provided by AI can help identify potential biases or manipulations. DeGrippo emphasized that those engaging in unscrupulous schemes typically do not employ overly complex strategies; they do just enough to meet their goals and not more.

The rapid advancement of AI technologies suggests that such manipulative behaviors will likely increase in frequency as both consumers and developers seek ways to navigate and mitigate unwanted influences. Staying informed and proactive is essential for users aiming to protect themselves in an ever-evolving digital landscape.

Author: Alaina Yee, Senior Editor

With over 15 years in technology and video games journalism, the author covers a wide range of topics, focusing primarily on security to help individuals understand online safety. Her expertise spans various subjects, including CPUs, Windows, and PC building.

Recent stories by the author:

• Microsoft warns about poisoned AI responses triggered by keywords.
• Trusting AI poses risks: A call for an open-source revival.
• Even cybersecurity professionals can make simple errors; here’s the lesson learned.

Source: PCWorld News