On March 23, SWE-rebench published its latest benchmark rankings, with Claude Opus 4.6 topping the list. This update removed example demonstrations and step limitations, while adding auxiliary evaluation tasks. GLM-5 ranked third among open-source models, with Chinese models collectively occupying four positions.

Baltimore City has filed a lawsuit against Elon Musk’s X Corporation and its AI tool Grok, alleging that they generated sexualized images without consent and involved content involving minors, violating the Consumer Protection Law. This case marks progress in local government efforts to advance AI regulation in the absence of federal legislation, and may impact future legal liability for AI companies.

Half of Norway's sovereign wealth fund employees are using Anthropic's Claude large language model to write AI tools to assist with decision-making and monitor investment risks. The fund's leadership stated that in the future, AI will be allowed to make limited decisions, but human oversight will still be required. The CEO emphasized the substantial returns from AI investments and plans to shift positions toward front-end investing.

Norway's sovereign wealth fund has approximately half of its employees using Anthropic's Claude large language model to create AI tools that assist with decision-making and analyzing ESG risk. In the future, the fund will permit AI to make autonomous decisions under supervision. The CEO stated that companies not using AI are "complete idiots." The fund has already seen significant returns on its AI investments.

ByteDance's cloud computing business Volcano Engine handles over 1 trillion tokens daily in large language model API calls on average, representing growth exceeding 60% compared to pre-Chinese New Year levels. The primary growth drivers stem from increased adoption among individual users and AI products, with rapid expansion occurring in both domestic and international markets. ByteDance plans to raise its Model-as-a-Service (MaaS) revenue targets to address intensifying competition.

Gate News reported that on March 25, Tencent announced the official launch of the desktop version of "Yuanbao Pai," its AI-native application Yuanbao. In the large-screen environment on desktop, users can participate in shared screen viewing while chatting and interacting with Pai friends or Yuanbao in a separate window. The desktop version of Yuanbao Pai supports multi-device message synchronization, file dragging, screenshot functionality, and other features, allowing users to complete information flow transfer within the platform.

OpenAI CEO Sam Altman announced that the new flagship model Spud has completed pretraining and is expected to be released soon. The company is restructuring, with the AI safety team transferred to the Chief Research Officer, the information security team moved to a new department, and Altman focusing on fundraising and data center construction. The product department has been renamed to "AGI Deployment," and multiple features will be consolidated into a desktop "super app."

OpenAI CEO Sam Altman announced at an all-hands meeting that the next-generation model Spud has completed pre-training and is expected to significantly accelerate the economy. The AI safety and information security teams have been reorganized, executive responsibilities adjusted, and plans are underway to integrate ChatGPT and other products into a desktop "super app."

According to 1M AI News monitoring, OpenAI COO Brad Lightcap stated at the Washington Hill and Valley Forum that memory chip shortages are the primary bottleneck for current AI infrastructure expansion: "Right now it's memory. In the past it was power." Tech companies are consuming increasingly more memory chip capacity, with large-scale purchases of Nvidia AI accelerators being the main driver. Memory shortages have spread from the AI industry to consumer electronics and automotive sectors.

Gate News reported that on March 25, OpenAI COO Brad Lightcap stated at the Hill and Valley Forum in Washington that memory chip shortages are the primary bottleneck for current AI infrastructure expansion: "Now it's memory. In the past, it was power." Tech companies are consuming increasingly more memory chip production capacity, with large-scale purchases of Nvidia AI accelerators being the main cause. Memory shortages have spread from the AI industry to consumer electronics and automotive sectors.