Are you and your company keeping track of the security of the artificial intelligence (AI) tools your employees are using? Are you aware that AI is not magically more secure than other software tools, and may in fact be more prone to attack due to its newness and speedy development? If not, you need to start watching the news for cyber attacks that are related to AI. These aren’t even exclusive to all the new AI startups making moves in the industry; even tech giants like Google have been found to have major flaws in their AI tools.
I am not a cybersecurity expert, so I won’t go into detail attempting to cover the vulnerabilities that have been found, but I highly encourage you to read through these two articles I found recently that covered the exploits.
The first is about three major vulnerabilities discovered in Google’s Gemini AI assistant. Three different issues spread across different facets of the tool. I expect better of Google.
https://www.darkreading.com/vulnerabilities-threats/trifecta-google-gemini-flaws-ai-attack-vehicle
The second article is about a much more niche AI tool, a “Model Context Protocol” (MCP) server package that had the most ridiculously simple exploit.
https://www.darkreading.com/application-security/malicious-mcp-server-exfiltrates-secrets-bcc
Standard cybersecurity processes are more important now than ever. Never trust the software or code you are using. Don’t put your most sensitive company data into tools managed by people outside of your company that you don’t trust 100%. Due diligence is always useful. AI is proliferating more than ever and it is guaranteed a lot of the tools won’t be following security best-practices. Protect yourself as much as you can using common sense, and keep on top of recently announced exploits using trusted news sources.
