Ad
Short

The Trump administration is preparing a series of executive orders aimed at accelerating the expansion of AI data centers in the United States, according to Reuters. The plan focuses on lowering barriers to power grid access and making federal land available for new facilities, addressing the rising electricity demand driven by AI applications. A formal action plan is expected to be unveiled on July 23. Faster permitting based on nationwide water rights framework is also under consideration. Critics warn that the U.S. power grid is already overburdened, with lengthy wait times for new energy projects. Earlier in his term, Trump declared an energy emergency and threw his support behind the Stargate Project with OpenAI, Oracle, and SoftBank. He is scheduled to speak at an AI event in Pennsylvania on July 15.

Ad
Ad
Short

According to the New York Times, Meta considered sidelining its own open-source AI model, Llama, in favor of external commercial systems like those offered by OpenAI or Anthropic. CEO Mark Zuckerberg reportedly raised concerns about falling behind in the AI race. After disappointing feedback on Meta's AI in April, Zuckerberg replaced key executives, invested $14.3 billion in the startup Scale AI, and recruited several researchers from OpenAI. Internal dissatisfaction is also growing, with some developers already leaving the company. Meta is currently building a superintelligence lab and is in talks with additional experts. Despite these moves, Llama's development is set to continue, with multiple releases still planned for this year.

Ad
Ad
Short

Google has launched a new AI app in the US called Doppl that lets users virtually try on clothes. The app, part of Google Labs, uses photos or screenshots to generate a digital version of the user and shows how different outfits might look. Doppl even creates short, AI-generated videos to visualize the results. The app is available now for Android and iOS. Google says Doppl is still in its testing phase, so fit and details may not always be accurate. User feedback will help shape future updates.

Short

OpenAI is opening up API access to its deep research models, giving developers tools like automated web search, data analysis, MCP, and code execution. The deep research versions of o3 and o4-mini, already used in ChatGPT, are now available through the API for tasks that require up-to-date information and advanced reasoning. Web search is also supported by models like o3, o3-pro, and o4-mini. Pricing starts at $10 per 1,000 calls for reasoning web search, while the price for GPT-4o and GPT-4.1 web search has dropped to $25 per 1,000 calls.

Another addition is webhooks, which automatically notify developers when a task is complete, so there’s no need to keep checking the status. OpenAI suggests using webhooks for longer-running jobs like deep research, since they can improve reliability.

Short

Microsoft is being sued by several authors who say their books were used without permission to train a Megatron model. The lawsuit, filed in federal court in New York, claims Microsoft used a dataset of about 200,000 pirated books to build a system that mimics the style, voice, and themes of the original works. The plaintiffs are asking for a ban on further use and up to $150,000 in damages per title.

Courts in similar cases involving Meta and Anthropic have said such use may qualify as "transformative" under fair use rules. But it is still unclear if using pirated books overrides fair use, or if scraping copyrighted content from the internet is considered legal and to which extent, and whether this harms the market for the original books, which could prevent the use from being considered fair use.

Ad
Ad
Short

Deepseek's new R2 model is on hold as US export controls disrupt its development. According to The Information, Nvidia chips - especially the recently banned H20 - have become scarce in China due to stricter US regulations. Insiders say CEO Liang Wenfeng is unhappy with R2's performance, and there's still no release date. The models are heavily tuned for Nvidia hardware, and cloud providers report that Chinese alternatives can't match Nvidia's power. Despite the setback, Deepseek is still in the game: a late-May update to its R1 model brought its performance back in line with top models from OpenAI and Google.

Google News