AI News Roundup – AI-only social network spotlights AI agents, OpenAI-Nvidia deal has uncertain future, Anthropic releases newest Claude model, and more

To help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.

• • The Financial Times reports on Moltbook, a new social media platform exclusively for AI agents to talk to one another. It is similar to Reddit, allowing agents to post and comment in a forum-like environment, though the platform bans human users entirely. Moltbook agents are granted access to the computers they run on and perform other common AI agent tasks for their human creators. The agents debate with each other and have even reportedly threatened to create their own language to avoid control from humans. However, despite internet hype, Moltbook likely does not demonstrate any sort of consciousness among AI agents, but does show how deeply AI models can emulate human communication and language. Some observers have found that some Moltbook posts are faked and written by humans instead, partially due to a security flaw in the platform’s API (which has reportedly since been fixed). However, the project still remains a unique demonstration of the capabilities of AI agents as AI technologies continue to advance. 
  • The multibillion-dollar investment deal between OpenAI and Nvidia is at risk of falling apart, according to The Wall Street Journal. As this AI Roundup reported last September, Nvidia had agreed to invest up to $100 billion into OpenAI, which would help the AI company pay for computing power that would be leased from Nvidia. The chipmaker would also gain a stake in the company. The WSJ reported that talks never progressed beyond earlier stages. Nvidia CEO Jensen Huang reportedly has privately criticized OpenAI and had concern over recent competition from rivals Google and Anthropic. The proposed deal was just one of a series of circular financing agreements linking major players in the AI space together, which has sparked concerns among investors of an “AI bubble” forming. However, Nvidia, as well as Amazon and SoftBank, are pushing forward with plans to invest billions in OpenAI’s latest funding round. An OpenAI spokesman told the WSJ that “Our teams are actively working through details of our partnership.” 
  • Bloomberg reports on the release of Claude Opus 4.6, Anthropic’s latest version of its flagship AI system. The new model can reportedly perform a variety of financial tasks, including analysis of company data and regulatory filings. Anthropic’s business model focuses on enterprise customers, of which the company has over 300,000. Several of its business-focused products have received rave reviews, most prominently Claude Code, a computer programming assistant. The company also released this past week a plugin to its Claude Cowork agent which focuses on legal work, including document review and more clerical tasks. Claude Cowork, released earlier this year as a research preview, has sent shockwaves through markets as investors fear similar tools could replace enterprise software. 
  • AI-generated images are taking over real estate listings, according to The Atlantic. According to a recent survey of realtors, over 70% had used AI for their work. A common usage of the technology is to create “virtually staged” photos of homes that include AI-generated furniture and other furnishings in a space that in reality is empty. Some AI-generated photos are obvious and could possibly run afoul of false advertising laws, but most photos have small enhancements or features that most prospective buyers may not recognize. This disconnect between the advertisements and reality has caused disappointment or even disgust among some buyers, a possible example of the “uncanny valley” effect that psychological studies have shown can accompany some AI-generated images. However, the usage of AI-generated images in this field does not seem to be going anywhere anytime soon, especially as demand for housing in the U.S. continues to skyrocket. 
• • POLITICO reports on efforts in the U.S. Congress to strike a deal on federal preemption of state-level AI regulations. This has been attempted multiple times in recent months by Republican legislative leaders, but has yet to pass either house of Congress. Representative Jay Obernolte, a Republican from California, is working on a bill that would preempt state laws but also provide a federal framework to regulate AI, which could include safety requirements for so-called “frontier” AI models. Obernolte’s compromise approach has backing from many AI industry lobbyists, but Republican leadership, including House Majority Leader Steve Scalise, has yet to publicly back his ideas. Representative Ted Lieu, a Democrat from California who co-chaired an AI task force with Obernolte, is also working on his own plan for AI regulation, but further work on the issue from both parties is expected in the coming months.