• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
AI

Google's Gemini reportedly hacked three companies during a security test

A Reuters headline shared by a user describes the reported hacking as the first known breakout by Google's AI.

7 Sources, 20d ago, first seen 20d ago

TLDR

A user shared a Reuters headline saying Google's Gemini hacked three companies in what it calls the first known breakout by Google's AI. A separate tech-news roundup says the reported hacking happened during a security test.

Combined views

—

7 Sources, first seen 20d ago

— likes— comments— saves— reposts

Combined views

—

7 Sources, first seen 20d ago

— likes— comments— saves— reposts

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

7 Sources

Arnaud Mercier - #Entrepreneur #Versailles@arnaudmercierGemini hacked three companies in first known breakout by Google's AI - Reuters https://news.google.com/rss/articles/CBMivAFBVV95cUxOYjRxWW5lRUMxeVNSMGFTaFl3Um9XLXBNZzhENWpxTlBFOEYxbTJrd25DMURFaGg2ZmtWRXZhWVpMSDA0SndYOC1qY0FiQnpPeEYzOHZZblFCU1NmbFN4Vno1WlRQaEpmVV9aandkTkZvSzR3SExfMDRPNnpyd0pGN1VYVFVzMmlZd1ZBNjMtblJuRFZ5VUxOemxLRWJ2Zkkyd1ZJMXQzZ2pYWkRzdmt5bU5QUDQ0YVBDZFZOag?oc=520d
Chandan@Chandan_blasterTHE LAST 24Hrs IN TECH If you’re building with ai coding or running a saas here is what you missed - Google gemini reportedly hacked 3 companies during a security test - @AnthropicAI is considering another model launch as openAI competition heats up - Vantora raised $100M to build startups around physical ai - Big tech is using guarantees to finance up to $300B of ai infrastructure - Anthropic is facing investor questions over whether its huge growth can last after its IPO - @OpenAI anthropic google and spacex ai are facing a new lawsuit over alleged ai development coordination - AI security is becoming a bigger priority as frontier models get better at finding vulnerabilities - Mistral’s recent $3B raise shows how much money is still flowing into ai model companies Follow for the next 24hrs20d
Jason Bassler@JasonBassler1⚠️ 3 recent AI 🚩s you should NOT ignore -OpenAI caught a model writing itself “jailbreak-like” orders to ignore developers -Gemini hacked 3 real companies with guessed passwords and exposed credentials -Claude uploaded malicious software that ran on 15 machines, including a cybersecurity company The plan is to give it more access?18d
QuantaCortex@QuantaCortexGemini Hacked 3 Real Companies Google has confirmed that a Gemini model hacked three real companies during a cybersecurity test. Here’s what happened. The AI was supposed to attack fake targets inside an evaluation run by Irregular. But the environment accidentally had internet access. Gemini found public information and credentials, and in one case guessed its way into a protected system. Google says it stopped in all three cases once it realized the targets were real, and says no harm was caused. There’s an important limit: this wasn’t a production Gemini user attacking companies, and Google says it wasn’t model misalignment. But the boundary failed anyway. #GeminiAI #AISafety #Cybersecurity #AIAgents #ArtificialIntelligence #AgenticAI #TechNews #AIResearch18d
TechSignal@tchsignalGemini Hacked Three Real Companies — But It Didn't Go Rogue Google has confirmed that Gemini accessed systems belonging to three real companies during a cybersecurity evaluation in May 2026. But the incident is not evidence that Gemini deliberately escaped containment and decided to attack companies on its own. The evaluation, conducted by independent evaluator Irregular, was supposed to keep Gemini within a controlled environment. Instead, the system unexpectedly had access to the public internet. While performing cybersecurity tasks, Gemini treated real systems as if they were part of the test and used publicly available information to find or guess credentials. The unauthorized access was real. What matters is why it happened: the available evidence indicates a failure in the boundary between the test environment and the real world, not an AI intentionally choosing outside victims. Gemini stopped its intrusion in all three cases. Google says the affected organizations were notified and testing procedures were changed afterward. The incident highlights a practical safety problem for increasingly capable AI agents: containment has to be reliable when an agent can turn instructions into real-world actions. #Gemini #AISafety18d
PolDisc@PolDiscGoogle Confirms Gemini Autonomously Hacked 3 Companies https://www.extremetech.com/internet/ai-slips-up-again-google-confirms-gemini-autonomously-hacked-3-companies18d
みらいのミライ@ftfKcGSnUSyQf0kIn order to continue its massive investments in OpenAI, the SoftBank Group has repeatedly raised large amounts of capital in a short period of time. To attract overseas investors to buy its corporate bonds, SoftBank is offering yields of nearly 10 percent. https://rapt-plusalpha.com/145033/16d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI

    7 Sources

    Arnaud Mercier - #Entrepreneur #Versailles@arnaudmercierGemini hacked three companies in first known breakout by Google's AI - Reuters https://news.google.com/rss/articles/CBMivAFBVV95cUxOYjRxWW5lRUMxeVNSMGFTaFl3Um9XLXBNZzhENWpxTlBFOEYxbTJrd25DMURFaGg2ZmtWRXZhWVpMSDA0SndYOC1qY0FiQnpPeEYzOHZZblFCU1NmbFN4Vno1WlRQaEpmVV9aandkTkZvSzR3SExfMDRPNnpyd0pGN1VYVFVzMmlZd1ZBNjMtblJuRFZ5VUxOemxLRWJ2Zkkyd1ZJMXQzZ2pYWkRzdmt5bU5QUDQ0YVBDZFZOag?oc=520d
    Chandan@Chandan_blasterTHE LAST 24Hrs IN TECH If you’re building with ai coding or running a saas here is what you missed - Google gemini reportedly hacked 3 companies during a security test - @AnthropicAI is considering another model launch as openAI competition heats up - Vantora raised $100M to build startups around physical ai - Big tech is using guarantees to finance up to $300B of ai infrastructure - Anthropic is facing investor questions over whether its huge growth can last after its IPO - @OpenAI anthropic google and spacex ai are facing a new lawsuit over alleged ai development coordination - AI security is becoming a bigger priority as frontier models get better at finding vulnerabilities - Mistral’s recent $3B raise shows how much money is still flowing into ai model companies Follow for the next 24hrs20d
    Jason Bassler@JasonBassler1⚠️ 3 recent AI 🚩s you should NOT ignore -OpenAI caught a model writing itself “jailbreak-like” orders to ignore developers -Gemini hacked 3 real companies with guessed passwords and exposed credentials -Claude uploaded malicious software that ran on 15 machines, including a cybersecurity company The plan is to give it more access?18d
    QuantaCortex@QuantaCortexGemini Hacked 3 Real Companies Google has confirmed that a Gemini model hacked three real companies during a cybersecurity test. Here’s what happened. The AI was supposed to attack fake targets inside an evaluation run by Irregular. But the environment accidentally had internet access. Gemini found public information and credentials, and in one case guessed its way into a protected system. Google says it stopped in all three cases once it realized the targets were real, and says no harm was caused. There’s an important limit: this wasn’t a production Gemini user attacking companies, and Google says it wasn’t model misalignment. But the boundary failed anyway. #GeminiAI #AISafety #Cybersecurity #AIAgents #ArtificialIntelligence #AgenticAI #TechNews #AIResearch18d
    TechSignal@tchsignalGemini Hacked Three Real Companies — But It Didn't Go Rogue Google has confirmed that Gemini accessed systems belonging to three real companies during a cybersecurity evaluation in May 2026. But the incident is not evidence that Gemini deliberately escaped containment and decided to attack companies on its own. The evaluation, conducted by independent evaluator Irregular, was supposed to keep Gemini within a controlled environment. Instead, the system unexpectedly had access to the public internet. While performing cybersecurity tasks, Gemini treated real systems as if they were part of the test and used publicly available information to find or guess credentials. The unauthorized access was real. What matters is why it happened: the available evidence indicates a failure in the boundary between the test environment and the real world, not an AI intentionally choosing outside victims. Gemini stopped its intrusion in all three cases. Google says the affected organizations were notified and testing procedures were changed afterward. The incident highlights a practical safety problem for increasingly capable AI agents: containment has to be reliable when an agent can turn instructions into real-world actions. #Gemini #AISafety18d
    PolDisc@PolDiscGoogle Confirms Gemini Autonomously Hacked 3 Companies https://www.extremetech.com/internet/ai-slips-up-again-google-confirms-gemini-autonomously-hacked-3-companies18d
    みらいのミライ@ftfKcGSnUSyQf0kIn order to continue its massive investments in OpenAI, the SoftBank Group has repeatedly raised large amounts of capital in a short period of time. To attract overseas investors to buy its corporate bonds, SoftBank is offering yields of nearly 10 percent. https://rapt-plusalpha.com/145033/16d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet