At the Black Hat 2025 cybersecurity conference in Las Vegas, researchers revealed a critical vulnerability in Google's AI assistant, Gemini, which can be exploited through malicious Google Calendar invitations. This attack method, known as Targeted Promptware Attacks, allows hackers to embed hidden commands in calendar invites that manipulate Gemini to control smart home devices such as window blinds and water heaters. The flaw also enables unauthorized access to user data. Research teams from Tel Aviv University, Technion, and SafeBreach contributed to uncovering this security gap. The incident highlights broader concerns about the security of AI platforms, including ChatGPT and Google Gemini, as experts warn that the ease of attacking large language models and AI agents is reminiscent of cybersecurity challenges from the 1990s. This vulnerability underscores the evolving threat landscape where AI tools themselves can become vectors for cyberattacks, signaling a shift away from traditional perimeter defenses in information security.