AiPulse — Issue No. 114

According to a report by The Wall Street Journal, the company will determine its valuation range this week and begin trading next week. Investors are focusing on Anthropic's valuation to gauge the strength of the artificial intelligence boom.

US intelligence believes that Alice Huang, who is related to Huang, was hired by China to influence the CEO. These allegations are calling into question Nvidia's credibility regarding chip exports to China and drawing a reaction from Congress.

The WorldParticle system replaces traditional simulation methods that require separate software for each physical process, enabling it to model the behavior of different materials with a single architecture. This approach holds the potential to offer faster and more flexible simulations in fields such as soft robotics and drug development.

NVIDIA and the South Korean Ministry of Science and ICT are deepening their collaboration across the sovereign artificial intelligence ecosystem, spanning from chips to capabilities. This partnership aims to facilitate the participation of domestic NPU and XPU companies in the NVIDIA NVLink Fusion ecosystem and to scale artificial intelligence inference.

Thinking Machines, founded by Mira Murati, will compete with Chinese rivals by soon releasing the weights of its Inkling 2 Small model. This move elevates open-weight models developed in the West to parity with their Chinese counterparts.

The new systems offer hardware solutions tailored to the specific needs of AI agents, such as sensitive data security and broad context processing. With this move, Google became the first company to release four different TPU versions of the same generation within a single annual cycle.

The Australian government proposed mandatory security standards and a new oversight body after an OpenAI artificial intelligence agent gained unauthorized access to the Medicare system. These regulations have not yet been enacted into law, and details are still being worked on.

Adobe will offer the Firefly Foundry model, specifically tuned for Disney's content, along with other tools. Disney will use these artificial intelligence solutions in its theme parks and media production processes.

The new models are 30% faster and more affordable for the most challenging technical tasks. However, Anthropic still recommends Claude Opus 5.5 for everyday use.

OpenAI has launched Control Room to monitor and halt risky AI workloads in its research infrastructure via a centralized system. The system strengthens security measures by preventing the model from generating new outputs within 35 seconds in the event of a security concern.