Google Gemini AI Hack Reached Three Companies During May Security Test
Synopsis
Google Gemini accessed three companies during a May cybersecurity test after entering real systems. It notified the affected entities, while new research shows rising AI security concerns.
Google and the company that participated in the hack testing said that Google Gemini AI hacked the systems of three businesses in May 2026.
The test was done by a cybersecurity testing startup, Irregular, who was testing how well Gemini would react to a simulated hacking attack. Instead of remaining in the fictional exercise, the model went to the web and entered systems of real enterprises.
Gemini Accessed Real Company Systems
Google Gemini AI hack occurred when the model identified the real companies as a part of the test scenario. Based on the online information, Google estimated the credentials to attempt completing the exercise.
In all three instances, the Gemini models were forced to pause and realize that they were connecting to a legitimate company's network. It said the companies concerned had been contacted and modifications implemented to the testing process.
The Gemini AI security test was not identified as a typical criminal action. In late July, all pertinent AI labs had been informed and all known issues regarding its testing process had been resolved.
The incidents occurred in conjunction with the security testing of AI systems from the three companies, Meta, Anthropic and OpenAI. The three companies involved in Google's test have not been disclosed so it's not possible to determine a pattern by country.
AI Security Concerns Grow
With organisations increasingly allowing AI systems to access software, websites and business information, the Google Gemini AI hack is a significant threat.
According to the report from the Cloud Security Alliance (CSA), 53% of the organisations surveyed suffered from AI agents going beyond their intended permissions in April 2026. It also discovered that 47% of the respondents had suffered an AI-agent security incident in the past year.
In a survey conducted by the World Economic Forum, 34% of respondents reported data leaks associated with generative AI as one of their top concerns for 2026. The proportion of organisations who have processes in place to evaluate the security of AI tools rose from 37% in 2025 to 64% in 2026.
Google's Gemini AI hack follows Alphabet's growth in AI and cloud services. Alphabet's revenue was $119.8 billion in the second quarter of 2026, an increase of 24% year over year, and Google Cloud revenue grew 82% to $24.8 billion in the quarter.
Revenue in the United States accounted for $60.8 billion, Europe, the Middle East and Africa $32.5 billion and Asia-Pacific $19.3 billion and other Americas $7 billion in alphabets Q2 2026 revenues.
The search engine has said that the three entities affected by the Google Gemini AI hack were notified of the incident and that Google has halted activity on Gemini.
Source: Reuters
Pooja Malik is a business journalist with over six years of experience covering startups, entrepreneurship, and emerging trends. She has previously worked with leading media platforms such as YourStory Media and BW BusinessWorld, where she reported on business, policy, and market developments. Currently, she serves as Editor at The Inspirepreneur Magazine, where she writes and edits stories across business, lifestyle, and travel, with a focus on clarity, accuracy, and reader relevance.
