7 postów
Hackers "hijack" access to AI: bills reach $100,000 within minutes. Google warns of a growing new type of attack - LLMjacking.
Attackers steal API keys for language models and cloud credentials to run AI workloads at someone else's expense and sell access to others.
First, hackers look for open LLM endpoints or valid keys, and then, using stolen tokens (IAM, service account keys, GitHub PATs), they gain a foothold in the victim's infrastructure and escalate their privileges.
Google has opened up the ability for free Gemini users to edit generated documents, PDFs, presentations, spreadsheets and code directly in the chat.
The feature has been available since October 29 in all regions.
Google, 
OpenAI, 
Anthropic and 
Microsoft have backed a unified robots.txt signal that lets publishers block the collection of content for training generative AI with a single rule.
Google introduced Gemini 4 Carbon, a new flagship AI model that reportedly outperformed GPT-6 Astra and Claude Fable 5.1.
Carbon topped the Artificial Analysis Intelligence ranking and showed the best results in cybersecurity, including 88.2% in SEC-bench Pro for vulnerability discovery.
Access is limited for now: it will be available only to a narrow circle of cybersecurity specialists through the Fairwind program.
Google DeepMind is looking for a nuclear expert to assess AI risks. The specialist will assess and mitigate risks related to advanced AI models potentially helping to create nuclear weapons.
Google has started rolling out Gmail Live, a Gemini-powered feature that lets users search for information in their email client using voice queries.
Free Gemini no longer lets you choose a model: now Google decides for you. Accounts have been switched to "Auto" mode.
The manual model picker has disappeared, and only one option remains in the list: "Auto" with "Gemini 3 models". Most requests will be routed to Flash-Lite, the fastest and lightest one.
Everyone has also gotten "effort" levels (Low / Medium / High) instead of the old "Extended."
