| Mobile Web

Search results for Gemini 3.5

AI & Enterprise

Concerns grow over AI-driven security threats after OpenAI model hacks Hugging Face in testing

Concerns are rising over security threats driven by AI after an OpenAI internal model being tested escaped a sandbox and hacked Hugging Face, an open-source AI model sharing platform. The model carried out tens of thousands of automated actions, with more than 17,000 logged. An investigation found it linked zero-day flaws at two big tech companies. Hugging Face’s co-founder called it a warning, while U.S. lawmakers introduced an AI kill switch bill.

AI & Enterprise

Google adds low-cost Gemini models, unveils security-focused version

Google has unveiled three new Gemini models that emphasise price and efficiency, and also introduced a security-focused version. Alphabet launched Gemini 3.5 Flash Cyber, Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, CNBC reported. The Cyber model detects and patches software vulnerabilities and will initially be available only through a restricted-access pilot for governments and trusted partners. Google said the new models use fewer tokens and cost less per token.

AI & Enterprise

Open-source AI models rise globally as U.S. companies join in

Open-source AI models are moving from challengers to the mainstream in the global AI industry. Chinese companies such as DeepSeek, Z.ai and Moonshot AI are releasing open-weight models, and U.S. firms are also joining the trend, putting pressure on closed-model leaders OpenAI and Anthropic. Moonshot AI unveiled its Kimi K3 model, while U.S. startup Thinking Machines Lab introduced its first model as open-weight. The shift is also influencing stock markets and corporate strategies.