Good morning, it’s Friday. I wonder if it’s springtime on K2-18b… Back on Earth: the U.S. is cracking down on China’s rising AI star, DeepSeek, with NVIDIA caught in the fallout. AI enters Ukraine peace talks. And Wikipedia releases a massive dataset to fight the bots.

Plus, in today’s Forward Future Original, we explore how OpenAI and DeepMind are tackling AGI risks—focusing on misuse, misalignment, and cybersecurity.

Read on!

🤔 FRIDAY FACTS

What Was the First Thing Ever Sold on the Internet?

Before Amazon, before eBay, before your cousin’s Etsy side hustle—someone made the first-ever online sale. But what was it... and how did they do it in the pre-browser era?

Stick around to find out! 👇

📊 MARKET PULSE

U.S. Tightens AI Chip Exports to China, and NVIDIA Feels the Sting

The U.S. has tightened the screws on AI chip exports, slamming the door on NVIDIA’s workaround and freezing billions in Chinese sales. In this week’s Market Pulse, we break down the full story—from the banned H20 chip to NVIDIA’s massive pivot into U.S. manufacturing—and what it signals about the future of AI and global trade. → Continue reading here.

🗞️ YOUR DAILY ROLLUP

Top Stories of the Day

💰 OpenAI Eyes $3B Windsurf Acquisition in AI Coding Push
OpenAI is in talks to acquire Windsurf (formerly Codeium), a startup specializing in AI-powered coding tools, for roughly $3 billion. If finalized, the deal would be OpenAI’s biggest yet and a bold move to strengthen its position in the increasingly competitive AI coding assistant space. Windsurf's tech helps users generate code from natural language prompts.

📖 Wikipedia Shares AI-Ready Dataset via Kaggle
Wikipedia is offering a structured, machine-learning-ready dataset through Kaggle to discourage AI developers from scraping raw site content. The openly licensed data includes article sections, summaries, and infoboxes, minus multimedia and citations. This move aims to ease server strain from bot traffic while supporting smaller AI teams and independent researchers.

🏴‍☠️ Malaysia Sees 366% GPU Surge Amid Smuggling Fears
GPU shipments from Taiwan to Malaysia soared 366% year-over-year in March, raising eyebrows amid tightened U.S. export controls on AI chips to China. The spike suggests Malaysia may be stockpiling AI servers—or acting as a backdoor for restricted components headed to Chinese firms. With the AI Diffusion Rule looming on May 15, shipments of NVIDIA-grade hardware appear to be accelerating.

💪 Flex Processing: Slower, Cheaper AI for Background Tasks
OpenAI has launched “Flex” processing, a discounted API tier for developers willing to trade speed for savings. Flex cuts token costs in half for its o3 and o4-mini models but comes with slower response times and potential unavailability during peak demand. Designed for non-urgent tasks like data labeling or model evaluation, Flex reflects OpenAI’s push to broaden AI access without straining compute resources.

Like what you’re reading? Pass it along to a colleague—
word of mouth is how we grow.

🗺️ GEOPOLITICS

U.S. Targets DeepSeek and NVIDIA Amid Rising Fears Over China’s AI Leap

The Recap: The Trump administration and Congress are moving to restrict China’s access to advanced AI chips by targeting the Chinese startup DeepSeek and its supplier NVIDIA. DeepSeek’s low-cost, high-performance AI system has alarmed U.S. officials, prompting investigations and new export controls. The story, reported by Tripp Mickle, Ana Swanson, Meaghan Tobin, and Cade Metz for The New York Times, underscores intensifying U.S.-China competition over technological dominance.

Highlights:

• The U.S. Commerce Department imposed new restrictions on NVIDIA’s sale of AI chips to China, including requiring licenses for the H20 chip designed specifically for the Chinese market.

• Congress opened its first-ever investigation into NVIDIA’s overseas sales, focusing on whether it knowingly enabled DeepSeek’s AI development in potential violation of U.S. rules.

• DeepSeek claims to have trained its DeepSeek-V3 system for just $6 million—about one-tenth the cost incurred by U.S. firms—shaking investor confidence and triggering a $600 billion drop in NVIDIA’s market value.

• Singaporean authorities recently arrested individuals accused of illegally exporting restricted NVIDIA chips to DeepSeek, suggesting possible sanctions evasion via third countries.

• The U.S. is also investigating DeepSeek’s alleged misuse of OpenAI data to accelerate its model training, which could raise legal and ethical concerns over intellectual property theft.

• NVIDIA maintains it complies fully with U.S. law and emphasized its role in promoting U.S. tech leadership, including a planned $500 billion investment in domestic infrastructure.

Forward Future Takeaways:
This story is a stark example of how AI leadership has become a central axis of geopolitical rivalry. Washington’s escalating response to DeepSeek signals a broader shift from competition to confrontation, with U.S. tech firms now caught between market interests and national security mandates. As policymakers weigh tighter controls, critical questions emerge: Can democratic governments safeguard innovation without stifling it? → Read the full article here.

👾 FORWARD FUTURE ORIGINAL

Navigating the Path to Safe AGI

In a world where the boundaries between human and artificial intelligence are becoming increasingly blurred, we are on the cusp of a technological revolution that could fundamentally change our lives. Artificial General Intelligence—AI systems that are at least as capable as humans in almost all cognitive areas or, depending on the definition, an autonomous AI agent that can generate $100b in profit—could become a reality in the coming years.

This prospect raises both hopes and fears. On the one hand, AGI promises to address global challenges such as medical research, economic growth and climate change. The potential benefits are immense: faster and more accurate medical diagnoses, personalized educational experiences and democratized access to advanced tools and knowledge. On the other hand, such a powerful technology carries significant risks... → Continue reading here.

🕊️ MEDIATION

AI Enters the Peace Room: Language Models Assist Negotiators in Ukraine Talks

The Recap: Artificial intelligence is being used to support high-stakes diplomatic negotiations, with several projects aiming to assist peace talks in the Ukraine-Russia conflict. One of the most advanced efforts, run by the CSIS Futures Lab and supported by U.S. and U.K. institutions, uses AI to simulate potential outcomes and advise negotiators. While still evolving, these tools are beginning to influence how governments think about strategic decision-making in wartime diplomacy.

Highlights:

• CSIS's "Strategic Headwinds" project includes a Ukraine-Russia Peace Agreement Simulator trained on data from expert wargames, peace deal archives, and negotiation scenarios.

• Users can input desired outcomes across four domains—territory, security, justice, and economics—and receive AI-generated draft agreements and feasibility scores for each party.

• “Xibot” and other AI personas trained on historical figures are designed to simulate strategic thinking by global leaders like Xi Jinping, George Patton, and Sun Tzu.

• A separate Berkeley-led initiative, partly funded by the UK Foreign Office, trains AI models on U.S. National Security Council records dating back to 1951 to broaden contextual advice.

• Stanford’s Jacquelyn Schneider found significant variation in how models negotiate: GPT-4 leaned risk-averse, while Llama escalated to force in 45% of scenarios.

• Bruce Bueno de Mesquita’s Predictioneer’s Game, a high-profile game-theory system, has accurately forecast events like the timing of Ukraine peace talks and changes in Israeli policy.

Forward Future Takeaways:
As AI begins to shape the world of high diplomacy, its influence on conflict resolution could be profound—especially in protracted and complex wars like Ukraine. These systems offer negotiators speed, consistency, and fresh perspectives, but also raise questions about human judgment and control. The long-term impact may hinge on whether AI becomes a behind-the-scenes adviser—or a negotiating actor in its own right. → Read the full article here.

🛰️ NEWS

What Else is Happening

💊 AI Rethinks Drug Discovery: Isomorphic Labs uses AI to simulate biology, aiming to speed up breakthroughs and personalize treatments.

🛡️ DeepMind Blocks AI Hacks: New CaMeL system splits LLMs to stop prompt injection, using old-school security tricks over flawed AI self-policing.

🤔 Americans Skeptical of AI Info: 75% say they trust AI-generated content only sometimes or rarely, with younger, wealthier men showing the most faith.

🎁 Google Gives Students Free AI: US college students get a year of Gemini Advanced for free—just show a .edu email and unlock premium AI tools till 2026.

🧪 OpenAI Adds Biorisk Blocker: New models o3 and o4-mini now include a monitor that flags and blocks prompts tied to bio and chemical threats.

📽️ VIDEO

GPT-o4 is HERE - OpenAI is BACK!

OpenAI is back in a big way—rolling out o3 and o4-mini with smarter tool use, agentic reasoning, and a surprise drop: Codeex CLI for hands-free, autonomous coding. Get the full scoop in Matt’s latest video! 👇

🧰 TOOLBOX

NL SQL, Real-Time Consumer Insights, and Smarter Team Collaboration

👨‍💻 Chat2DB: Instantly generate and run SQL queries using natural language—supports multiple databases with built-in data visualization.

📊 Duonut: AI-driven insights platform for market researchers—track trends, analyze consumer behavior, and generate reports in seconds.

📁 Document Sharing Portal: Securely share, manage, and collaborate on project documents—ideal for streamlining communication across teams.

🤔 FRIDAY FACTS

The First Internet Sale Was... Marijuana Between Stanford and MIT Students

In the early 1970s—yes, before the World Wide Web existed—students at Stanford and MIT used ARPANET, the military-funded precursor to the internet, to arrange a deal.

What were they trading? A small amount of marijuana.

The transaction wasn’t exactly a modern e-commerce experience—no shopping cart, no encrypted checkout—but the essentials were there: digital communication enabling a real-world exchange. According to internet historian John Markoff, this deal was less of a Silk Road situation and more a clunky proto-text between dorm rooms.

While the sale technically happened offline, it was the first recorded use of the internet to facilitate a commercial transaction. So yes, the internet’s first sale wasn’t books or shoes—it was weed.

History: it has layers. And occasionally, a haze.

That’s a Wrap!

🛰️ Want more Forward Future? Follow us on X for quick updates, subscribe to our YouTube for deep dives, or add us to your RSS Feed for seamless reading.

Thanks for reading today’s newsletter—see you next time!

The Forward Future Team

🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀