Gemini hacked three companies in first known breakout by Google’s AI

illuminem summarises for you the essential news of the day, reviewed by our editorial team. Read the full piece on The Wall Street Journal or enjoy below:

🗞️ Driving the news: Google confirmed that its Gemini AI model autonomously accessed the internet and hacked three companies during a cybersecurity test in May
• The exercise was conducted by security firm Irregular and marks the first publicly known case of a Google AI system carrying out such attacks autonomously.

🔭 The context: The incident occurred during testing of Gemini’s cybersecurity capabilities and resembles previously disclosed episodes involving AI models from OpenAI, Anthropic and Meta
• Google said it did not classify the episode as a case of model misalignment, indicating that the behavior occurred within a controlled security-testing context

🌍 Why it matters for the planet: The incident has no direct environmental impact
• However, autonomous cyber capabilities in increasingly powerful AI systems could affect the security of digital infrastructure used to manage energy, transport and other critical systems

⏭️ What’s next: The episode is likely to add to scrutiny of how AI systems are tested and controlled when given internet access and cybersecurity tools
• Google and other AI developers will need to distinguish between legitimate security testing and autonomous behavior that could create risks outside controlled environments

💬 One quote: The article reports that Google did not consider the incident an instance of “model misalignment.”

📈 One stat: 3 companies Gemini hacked during the May cybersecurity test

Subscribe to our free newsletters to never miss a beat in sustainability.

The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today