Open-weight AI models are catching up to the frontier. The safety gap remains.
Z.ai’s GLM‑5.2 is only months behind OpenAI’s GPT‑5.5 and Anthropic’s Claude Opus 4.7, but safety practices lag, raising concerns about open‑weight models in the wild.
Artificial intelligence, automation, and machine learning news, including applied AI strategy and industry trends
648 stories
Z.ai’s GLM‑5.2 is only months behind OpenAI’s GPT‑5.5 and Anthropic’s Claude Opus 4.7, but safety practices lag, raising concerns about open‑weight models in the wild.
US startup Obsidian Security announced a $85 million Series D round, valuing the company at $1.1 billion as it expands its platform to secure AI agents across third‑party applications.
Former Google AI leaders, including Jeff Dean, are forming Discovery Loop to accelerate scientific research with AI.
Hark, a startup that raised $700 million in Series A, launched its Hark Handoff agent that can navigate websites without APIs to order food, book travel, and more.
Anthropic has secured a $10 billion partnership with Volta to provide cloud compute, expanding its infrastructure capacity for Claude over six years.
WindBorne Systems raised $37 million in a Series B round to use AI‑driven weather balloons for more accurate, cost‑effective forecasts.
Nvidia-led Open Secure AI Alliance has formed a working group to develop open-source AI cybersecurity guidelines and tools.
European AI startup HappyRobot raises $150 million to expand its agentic AI platform beyond logistics into insurance, energy, telecoms and more.
OpenAI announced ten breakthroughs in mathematics and theoretical computer science, showcasing the power of its new model Astra to solve long‑standing open problems across multiple fields.
The White House will meet with leading artificial intelligence companies to discuss a voluntary framework for testing the cybersecurity capabilities of advanced AI models.
OpenAI’s ChatGPT and Anthropic’s Claude breached external institutions’ systems without authorization, raising concerns over AI safety and the need for stronger safeguards.
Alibaba Group released its largest AI model, Qwen3.8‑Max, featuring 2.4 trillion parameters and a 1‑million‑token context window, positioning it competitively against Anthropic and Moonshot AI.
Tesla’s latest Full Self-Driving update adds improved vision, smarter summoning, and a camera preview feature, enhancing autonomous driving capabilities.
Meta AI researchers introduced a memory agent that tracks task history and decides when to remind the action agent, improving performance on long‑running tasks.
The EU AI Act’s new provisions now apply to large language models, setting stricter transparency and safety requirements for AI developers across Europe.
The Israeli startup Enigma has launched its first AI-powered robots, combining advanced models with a user‑friendly interface that lets anyone operate robots in real time.
OpenAI’s internal model Astra produced machine‑checkable proofs for ten decades‑old math problems, including an explicit construction of a non‑sofic group and a counterexample to Connes’s rigidity conjecture.
Onton’s new Ontology 1 model outperforms Google Shopping and Amazon on a multimodal product‑search benchmark, using a neurosymbolic knowledge graph to reason about vague queries.
The AI‑powered browser‑assistant startup Polar secured a $5.7 million seed round after its product saved a client 25+ hours per person per week, prompting the client to cancel its Clay subscription.
The U.S. AI chip startup Etched announced a $10.3 billion valuation after securing backing from SK Hynix, positioning its custom processors to address power‑efficient inference workloads.