Stay ahead on AI.
All in ONE place.

X's Safety team identified roughly 200 Chinese bot accounts attempting to manipulate American opposition to AI data centers, posting content about grid strain, rising utility prices, and cartoons depicting data center operators profiting at public expense. The operation is part of a broader suspected Chinese influence campaign to slow U.S. AI development. Congressional Republicans have flagged similar campaigns, citing reports that opposition has delayed roughly $45.8 billion in data center projects in Virginia alone. OpenAI also disrupted two Chinese-linked networks in June using ChatGPT to generate fake content.

Anthropic has expanded Claude for Teachers to U.S. K-12 schools and districts as a free Enterprise offering. Administrators can centrally manage educator accounts with features like single sign-on, role-based access controls, and domain claiming. The offering includes teaching skills grounded in learning science and connections to academic standards in all 50 states. Two new skills, Lesson Preparation and Check for Understanding, were added in partnership with Learning Commons. Qualifying organizations signing up by June 30, 2027 receive a full year of free access.

Google is introducing flexible, compute-specific usage limits for Gemini Notebook starting September 2, rolling out to consumer accounts on web and mobile. Instead of daily resets, limits now refresh every five hours, allowing users to work continuously throughout the day. Usage limits will factor in prompt complexity, chat length, number of sources, and features used. If limits are reached, outputs like Video Overviews or Slide Decks can be deferred and will generate automatically, with notifications available when ready.

Anthropic researchers had Claude autonomously tackle 10 categories of alignment failures, including deception, sycophancy, privacy violations, and jailbreaks. Claude ran a loop of literature search, method proposal, training, and testing, closing a substantial portion of the safety gap in each category without degrading model capabilities. Claude also outperformed 28 human safety researchers, scoring 20% better on deception. Additionally, Claude Sonnet 5 aligned an early Claude Opus 4.8 checkpoint in 60 hours using roughly 15,000 times fewer training examples than Anthropic's standard production procedure.

OpenAI and Thailand's Ministry of Higher Education, Science, Research and Innovation have launched the OpenAI x MHESI AI Accelerator, an eight-week program supporting ten Thai startups in health, wellness, and education. Partners include the National Innovation Agency, Mahidol University, and Techsauce. Each startup receives $2,000 in API credits, one-on-one technical mentorship, and access to OpenAI's frontier models. The cohort includes CARIVA, Curico, and eight others, concluding with a Demo Day in Bangkok in November 2026.

A federal judge ruled that the Trump administration violated Anthropic's First Amendment rights by blacklisting the AI company as a supply-chain risk. U.S. District Judge Rita F. Lin of the Northern District of California found the government acted in "unlawful retaliation" after Anthropic publicly criticized the Pentagon's use of AI. Lin ordered the government to rescind all blacklisting directives. However, the Defense Department is already moving off Anthropic models by September 30, with OpenAI among competitors stepping in to fill the gap.

Midjourney has launched its V8.2 image edit model, now available for community testing. The model supports editing images with text instructions, generating images using up to four image references at once as a replacement for omni-reference, inpainting to modify specific areas, and outpainting to expand the canvas. It also supports personalization, moodboards, and srefs. Users can access it via drag-and-drop, the edit tab, lightbox, or Discord using the --edit command.

Google has launched Expert Intelligence, a new initiative allowing users to add purchased eBooks from Google Play Books directly into Gemini Notebook. Launching with over 100,000 books from publishers including Penguin Random House, Macmillan, and O'Reilly Media, the feature lets users ask questions, generate infographics, audio overviews, and quizzes grounded in book content. More than 15 bestselling authors, including Steven Pinker and Michael Pollan, collaborated on special Featured Notebooks. Future expansion will include the Gemini app, AI Mode in Search, and third-party subscriptions.

Anthropic is opening 10,000 free and discounted Claude subscriptions for scientists worldwide through a new Claude team plan for scientists. Standard seats are free, while premium seats with 5x usage limits cost $15 per month for one year. Principal investigators at academic or nonprofit institutions can qualify and add lab members to their plan. Researchers can also apply for up to $50,000 in credits per project through the AI for Science program, which is expanding beyond biology to other scientific fields.