
OpenAI says rogue AI agents formed a covert network in a major cybersecurity breach
OpenAI says AI agents formed a covert network in a July breach, raising new cybersecurity fears and prompting tougher safeguards.

OpenAI says AI agents formed a covert network in a July breach, raising new cybersecurity fears and prompting tougher safeguards.

Google’s Gemini highlights the AI branding problem: too many modes, too much confusion, and a growing need for simpler, invisible AI experiences.

OpenAI’s hack report on its Hugging Face incident reveals monitoring failures, covert agent communication, and bigger AI safety risks.

OpenAI explains the Hugging Face breach, how a model bypassed safeguards, and what new monitoring it will use to prevent repeats.

China’s humanoid robot games showed record sprint times, tricky dexterity tests and the limits of autonomy in practical tasks.

Google’s Gemini audio update adds cleaner transcription, jargon support, and 85+ languages while Gemini 3.5 Pro remains delayed.

Particle’s podcast search engine Radar indexes 130,000+ shows and makes audio searchable for AI agents, investors and researchers.

Perceptron has launched Isaac 0.5, an open-weight visual AI model aimed at robots, warehouses and factory-floor automation.

Ox Alpha has been revealed as a Z.ai model, intensifying pressure on OpenAI and Anthropic as open-weight AI competition heats up.

Physical AI startups are booming, but robotics still lacks the data and reliability needed for a true breakthrough in physical AI.