The #1 AI Newsletter for Business Leaders

Join 400,000+ executives and professionals who trust The AI Report for daily, practical AI updates.

Built for business—not engineers—this newsletter delivers expert prompts, real-world use cases, and decision-ready insights.

No hype. No jargon. Just results.

Subscribe free—trusted by leaders

Today:

• DeepSeek Model Hits Top Scores

• Sora Video AI Now Free

• Musk’s xAI Launches $5B Sale

• Character.AI Adds Video, Social Feed

• Apple Holds Back ChatGPT Rival

Deepseek just BROKE the Entire AI Industry... (something is up)

Deepseek’s May-28 R10528 upgrade vaults the open-source model to near-top test scores, matching OpenAI’s o3 and edging Gemini 2.5 Pro. Code and language “benchmarks” confirm the leap. 

Analysis shows the team shifted from training on OpenAI outputs to “synthetic” Gemini data, a shortcut called distillation. Deepseek’s pricing undercuts rivals, threatening their profit plans. The U.S.–China AI race intensifies as free, high-quality models erode closed competitors’ edge.

WATCH THE VIDEO ON YOUTUBE

Sora Video AI Now Free

Microsoft’s Bing mobile app now lets anyone use OpenAI’s Sora video model free. Users type a prompt—say, “tiny astronaut on a mushroom planet”—and the tool returns a five-second clip within seconds. After ten quick uses, clips render slower unless users spend reward points, Microsoft’s free credits. Landscape videos roll out worldwide soon except China and Russia. Built-in safety filters and labels mark AI output and block risky prompts.

Why this matters

1. Mass access, low cost – Making Sora free in Bing removes the paywall, putting advanced video generation in millions of pockets and widening real-world testing.
   

2. Rising competition – Microsoft’s move answers growing pressure from rivals such as Runway and Luma, showing how fast the generative-video race is evolving.
   

3. Enterprise signal – A simple tool that turns text into share-ready clips hints at future business uses, from quick training videos to automated marketing content.

Musk’s xAI Launches $5B Sale

Elon Musk’s AI firm xAI plans a $5 billion debt sale, arranged by Morgan Stanley, to bankroll massive data-center expansion and its planned “Colussus” supercomputer. The move follows xAI’s merger with social network X, valued together at $113 billion, and a possible $300 million employee share sale. Musk, now refocused on his companies, aims to add over a million AI chips and raise a further $20 billion from investors to fuel fast model growth.

Why it matters

1. Bigger war chest, bigger models – Raising debt shows how much cash today’s AI labs need for ever-larger computing power, hinting at even more capable systems ahead.
   

2. Tighter Musk ecosystem – Folding xAI and X together means social-media data can train xAI’s models, sharpening the debate over data ownership and privacy.
   

3. Intensifying competition – The funding push signals Musk’s intent to challenge OpenAI, Google and others, accelerating the race for cutting-edge AI.

Character.AI Adds Video, Social Feed

Character.AI is adding new tools that let users turn their chat characters into short videos and share them on a social feed. The AvatarFX feature creates five clips daily from a photo, chosen voice, and typed lines. Scenes supply preset stories; upcoming Streams pairs two characters live. The company blocks real-person photos and watermarks clips to limit deepfakes—fake videos—but safety worries persist after past harmful chatbot incidents, including self-harm suggestions.

Why this matters

1. Text-to-video for everyone – A playful consumer app shows how fast video generation is becoming a mainstream feature, pushing other AI platforms to follow.
   

2. Safety vs. creativity tension – New media powers arrive alongside risks like deepfakes and self-harm prompts, spotlighting the urgent need for stronger guardrails in generative AI.
   

3. Evolving social AI – Mixing chatbots, video, and public feeds hints at future social networks where AI characters become content creators and companions, reshaping online interaction.

🧠RESEARCH

ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models

Researchers show that training language models longer with reward-based practice lets them discover new ways to reason. Their ProRL method balances ideas with past knowledge, restarts learning often, and uses many tasks. The resulting models solve problems their originals cannot, and keep improving with more training time and extra data.

AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time

AlphaOne lets AI models switch between slow and quick thinking while answering. A single dial called alpha sets how long the model pauses to reflect. Slow-thinking steps add depth, then an end-thinking mark signals the final reply. Tests on math, code, and science tasks show faster answers with higher accuracy.

Time Blindness: Why Video-Language Models Can't See What Humans Can?

SpookyBench is a new test where each frame looks like snow. The only clue is how that noise shifts over time. People easily spot shapes and words in the moving specks, but top video-chat models score zero. The work shows systems ignore time patterns and need fresh designs to succeed.

🛠️TOP TOOLS

SEO GPT - AI-powered tool designed specifically for search engine optimization tasks. 

StockImg AI - AI-powered platform that revolutionizes visual content creation. 

Plazmapunk - AI-powered tool that transforms audio files into visually stunning music videos.

Boords - Designed to simplify and streamline the video production process. 

StudyX - AI-powered educational platform designed to provide comprehensive homework assistance and learning support for students.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Apple is testing powerful AI models matching ChatGPT’s level but isn’t releasing them yet. At WWDC 2025, it will offer smaller on-device models for basic features. Key projects like a smarter Siri are delayed until 2026.

• Microsoft will invest $400 million in Switzerland to boost its AI and cloud services. The money will expand data centers near Zurich and Geneva, support local businesses, and train workers in digital tools.

• Salesforce is bringing on part of Moonhub’s team after the startup announced its shutdown. Moonhub built AI tools to help companies hire talent. Though not a full acquisition, the move supports Salesforce’s growing AI strategy in HR tech.

• Snowflake plans to buy Crunchy Data to add its open-source Postgres database tech to the Snowflake AI Data Cloud. This move aims to help businesses build AI agents more easily using familiar tools. The deal supports regulated industries needing secure, compliant databases.

• The FDA has launched "Elsa," a generative AI tool built on AWS GovCloud to help staff review clinical protocols, summarize drug safety data, and generate code. It aims to speed up scientific evaluations but has raised concerns over data security and limited pilot testing so far.

• Thomson Reuters launched agentic AI tools starting with CoCounsel for tax and accounting, enabling AI to plan and act within workflows. Built with expert input, these tools boost accuracy, speed, and trust in high-stakes tasks.

• Laurene Powell Jobs supports Jony Ive’s secretive AI device built with OpenAI, calling his design process “wondrous.” Ive sees the project as a way to correct past tech harms and reclaim optimism in innovation.

• Educators across the U.S. say generative AI has disrupted teaching, grading, and learning. Many describe burnout, confusion, and distrust as students submit AI-written work. Some call it “demoralizing” and fear it’s eroding critical thinking altogether.